diff --git a/api/proto/luckygift/v1/luckygift.pb.go b/api/proto/luckygift/v1/luckygift.pb.go
index 1f532914..9ec48743 100644
--- a/api/proto/luckygift/v1/luckygift.pb.go
+++ b/api/proto/luckygift/v1/luckygift.pb.go
@@ -518,7 +518,7 @@ type LuckyGiftRuleConfig struct {
 	JackpotGlobalRtpMaxPpm    int64   `protobuf:"varint,28,opt,name=jackpot_global_rtp_max_ppm,json=jackpotGlobalRtpMaxPpm,proto3" json:"jackpot_global_rtp_max_ppm,omitempty"`
 	JackpotUserRoundRtpMaxPpm int64   `protobuf:"varint,29,opt,name=jackpot_user_round_rtp_max_ppm,json=jackpotUserRoundRtpMaxPpm,proto3" json:"jackpot_user_round_rtp_max_ppm,omitempty"`
 	JackpotUser_48HRtpMaxPpm  int64   `protobuf:"varint,30,opt,name=jackpot_user_48h_rtp_max_ppm,json=jackpotUser48hRtpMaxPpm,proto3" json:"jackpot_user_48h_rtp_max_ppm,omitempty"`
-	// 历史消费门槛字段只用于旧配置回读；dynamic_v3 当前大奖资格不再消费该值。
+	// 用户在同一应用、奖池和 UTC 自然日每累计消费满该金币数，生成一次可跨日保留的消费大奖资格。
 	JackpotSpendThresholdCoins int64 `protobuf:"varint,31,opt,name=jackpot_spend_threshold_coins,json=jackpotSpendThresholdCoins,proto3" json:"jackpot_spend_threshold_coins,omitempty"`
 	MaxJackpotHitsPerUserDay   int64 `protobuf:"varint,32,opt,name=max_jackpot_hits_per_user_day,json=maxJackpotHitsPerUserDay,proto3" json:"max_jackpot_hits_per_user_day,omitempty"`
 	MaxSinglePayout            int64 `protobuf:"varint,33,opt,name=max_single_payout,json=maxSinglePayout,proto3" json:"max_single_payout,omitempty"`
diff --git a/api/proto/luckygift/v1/luckygift.proto b/api/proto/luckygift/v1/luckygift.proto
index 1678ed8f..9f8c52c0 100644
--- a/api/proto/luckygift/v1/luckygift.proto
+++ b/api/proto/luckygift/v1/luckygift.proto
@@ -96,7 +96,7 @@ message LuckyGiftRuleConfig {
   int64 jackpot_global_rtp_max_ppm = 28;
   int64 jackpot_user_round_rtp_max_ppm = 29;
   int64 jackpot_user_48h_rtp_max_ppm = 30;
-  // 历史消费门槛字段只用于旧配置回读；dynamic_v3 当前大奖资格不再消费该值。
+  // 用户在同一应用、奖池和 UTC 自然日每累计消费满该金币数，生成一次可跨日保留的消费大奖资格。
   int64 jackpot_spend_threshold_coins = 31;
   int64 max_jackpot_hits_per_user_day = 32;
   int64 max_single_payout = 33;
diff --git a/services/lucky-gift-service/cmd/strategy-sim/main.go b/services/lucky-gift-service/cmd/strategy-sim/main.go
index b3f6f9fc..232c5a09 100644
--- a/services/lucky-gift-service/cmd/strategy-sim/main.go
+++ b/services/lucky-gift-service/cmd/strategy-sim/main.go
@@ -103,6 +103,7 @@ func runSimulation(options simulationOptions) simulationReport {
 		simulateSixRiskCapacities(config),
 		simulateRechargeStages(config),
 		simulateCombinationPriority(config, options.Seed+70),
+		simulateDualJackpotSpendStrategy(config, options.Seed+75),
 		simulateConcurrencyAndIdempotency(config, options.Seed+80),
 		simulateLongRunGroups(config, options.Seed+90, options.GroupLongRunN),
 	}
diff --git a/services/lucky-gift-service/cmd/strategy-sim/scenarios.go b/services/lucky-gift-service/cmd/strategy-sim/scenarios.go
index d1f81551..106a748b 100644
--- a/services/lucky-gift-service/cmd/strategy-sim/scenarios.go
+++ b/services/lucky-gift-service/cmd/strategy-sim/scenarios.go
@@ -938,6 +938,10 @@ func simulateCombinationPriority(config domain.StrategyConfig, seed int64) scena
 	config.HighWaterFactorPPM = 1_300_000
 	config.RechargeFactorPPM = 1_100_000
 	state := qualifiedSettledRoundState(10_000, 100*60*60*1000)
+	config.JackpotMechanism2Enabled = true
+	config.MilestoneSpendCoins = 50
+	state.PendingSpendJackpotTokens = 1
+	state.SpendGlobalRTP = domain.StrategyRTP{WagerCoins: 10_000, PayoutCoins: 9_800}
 	state.ConsecutiveZeroDraws = 5
 	input := domain.StrategyInput{
 		GiftPriceCoins: 10, NowMS: 1_000_000, HasRechargeFact: true,
@@ -947,8 +951,8 @@ func simulateCombinationPriority(config domain.StrategyConfig, seed int64) scena
 	if err != nil {
 		return failedScenario(name, err)
 	}
-	// 已结算轮次资格在普通概率、第6抽保底和动态因子之前尝试；独立消费里程碑机制已经停用。
-	priorityPass := priority.JackpotMechanism == domain.StrategyJackpotMechanismRTPCompensation && !priority.NextState.SettledRoundPending && priority.Stage == domain.StageAdvanced
+	// 两套资格同时存在时，48h 补偿先支付，消费 token 保留到下一个子抽。
+	priorityPass := priority.JackpotMechanism == domain.StrategyJackpotMechanismRTPCompensation && !priority.NextState.SettledRoundPending && priority.NextState.PendingSpendJackpotTokens == 1 && priority.Stage == domain.StageAdvanced
 
 	blockedState := state
 	blockedState.PoolBalanceCoins = 50
@@ -956,15 +960,118 @@ func simulateCombinationPriority(config domain.StrategyConfig, seed int64) scena
 	if err != nil {
 		return failedScenario(name, err)
 	}
-	// 大奖资格不能透支奖池；该轮资格被消费后，同一抽仍可按普通第6抽保底支付恰好可负担的5x。
-	blockedPass := !blocked.NextState.SettledRoundPending && blocked.JackpotMechanism == "" && blocked.PayoutCoins == 50 && blocked.PoolAfterCoins == 0
+	// 大奖资格不能透支奖池；48h 资格本轮结束，消费 token 则等待后续补水。
+	blockedPass := !blocked.NextState.SettledRoundPending && blocked.NextState.PendingSpendJackpotTokens == 1 && blocked.JackpotMechanism == "" && blocked.PayoutCoins == 50 && blocked.PoolAfterCoins == 0
 	return scenarioResult{
 		Name: name, Passed: priorityPass && blockedPass,
-		Summary: fmt.Sprintf("优先级=已结算轮次资格>普通P/W/第6抽；大奖池不足时资格消费，本抽仅赔可支付%s", blocked.SelectedTier.ID),
+		Summary: fmt.Sprintf("优先级=48h补偿>消费达标>普通P/W；48h成功不扣消费token，池不足时消费token保留，本抽仅赔%s", blocked.SelectedTier.ID),
 		Data:    map[string]any{"all_eligible": priority, "special_pool_blocked": blocked},
 	}
 }
 
+func simulateDualJackpotSpendStrategy(config domain.StrategyConfig, seed int64) scenarioResult {
+	const name = "双大奖消费资格矩阵"
+	config = settledRoundJackpotConfig(config)
+	config.JackpotMechanism2Enabled = true
+	config.MilestoneSpendCoins = 50
+	input := domain.StrategyInput{GiftPriceCoins: 10}
+	base := domain.StrategyState{
+		PoolBalanceCoins: 10_000, PendingSpendJackpotTokens: 1,
+		SpendGlobalRTP: domain.StrategyRTP{WagerCoins: 10_000, PayoutCoins: 9_800},
+	}
+	type row struct {
+		Case         string `json:"case"`
+		GlobalWager  int64  `json:"global_wager"`
+		GlobalPayout int64  `json:"global_payout"`
+		TokensBefore int64  `json:"tokens_before"`
+		TokensAfter  int64  `json:"tokens_after"`
+		Mechanism    string `json:"mechanism"`
+		Jackpot      bool   `json:"jackpot"`
+		Payout       int64  `json:"payout"`
+		PoolAfter    int64  `json:"pool_after"`
+		Reason       string `json:"reason"`
+	}
+	rows := make([]row, 0, 8)
+	run := func(caseName string, state domain.StrategyState, drawInput domain.StrategyInput, offset int64) (domain.StrategyDecision, error) {
+		decision, err := domain.DecideLuckyGiftStrategy(config, state, drawInput, domain.NewSeededStrategyRandom(seed+offset))
+		if err == nil {
+			rows = append(rows, row{
+				Case: caseName, GlobalWager: state.SpendGlobalRTP.WagerCoins, GlobalPayout: state.SpendGlobalRTP.PayoutCoins,
+				TokensBefore: state.PendingSpendJackpotTokens, TokensAfter: decision.NextState.PendingSpendJackpotTokens,
+				Mechanism: decision.JackpotMechanism, Jackpot: decision.Jackpot, Payout: decision.PayoutCoins,
+				PoolAfter: decision.PoolAfterCoins, Reason: decision.Trace.BlockedReason,
+			})
+		}
+		return decision, err
+	}
+
+	equal, err := run("大盘等于98%", base, input, 0)
+	if err != nil {
+		return failedScenario(name, err)
+	}
+	aboveState := base
+	aboveState.SpendGlobalRTP.PayoutCoins = 9_801
+	above, err := run("大盘高于98%", aboveState, input, 1)
+	if err != nil {
+		return failedScenario(name, err)
+	}
+	noSampleState := base
+	noSampleState.SpendGlobalRTP = domain.StrategyRTP{}
+	noSample, err := run("无已结算大盘", noSampleState, input, 2)
+	if err != nil {
+		return failedScenario(name, err)
+	}
+
+	// 40->150 crosses 50/100/150 in one paid child draw. A scripted zero proves those three new
+	// tokens cannot influence the current selection; the following child draw consumes only one.
+	crossState := base
+	crossState.PendingSpendJackpotTokens = 0
+	crossState.UserDaySpendCoins = 40
+	crossInput := domain.StrategyInput{GiftPriceCoins: 110}
+	crossed, err := domain.DecideLuckyGiftStrategy(config, crossState, crossInput, &scriptedRandom{indexes: []int64{0}, bounds: []int64{1_000_000}})
+	if err != nil {
+		return failedScenario(name, err)
+	}
+	rows = append(rows, row{Case: "当前子抽跨三档", GlobalWager: crossState.SpendGlobalRTP.WagerCoins, GlobalPayout: crossState.SpendGlobalRTP.PayoutCoins, TokensBefore: 0, TokensAfter: crossed.NextState.PendingSpendJackpotTokens, Mechanism: crossed.JackpotMechanism, Jackpot: crossed.Jackpot, Payout: crossed.PayoutCoins, PoolAfter: crossed.PoolAfterCoins, Reason: crossed.Trace.BlockedReason})
+	next, err := run("下一子抽消费一档", crossed.NextState, input, 3)
+	if err != nil {
+		return failedScenario(name, err)
+	}
+
+	qualified := qualifiedSettledRoundState(10_000, 100*60*60*1000)
+	qualified.PendingSpendJackpotTokens = 1
+	qualified.SpendGlobalRTP = base.SpendGlobalRTP
+	priority, err := run("双资格48h优先", qualified, input, 4)
+	if err != nil {
+		return failedScenario(name, err)
+	}
+	qualified.UserRoundRTP.PayoutCoins = 9_600
+	fallback, err := run("48h不合格消费接续", qualified, input, 5)
+	if err != nil {
+		return failedScenario(name, err)
+	}
+	poolBlockedState := base
+	poolBlockedState.PoolBalanceCoins = 49
+	poolBlocked, err := run("奖池不足资格保留", poolBlockedState, input, 6)
+	if err != nil {
+		return failedScenario(name, err)
+	}
+
+	passed := equal.JackpotMechanism == domain.StrategyJackpotMechanismMilestone && equal.NextState.PendingSpendJackpotTokens == 0 &&
+		!above.Jackpot && above.NextState.PendingSpendJackpotTokens == 1 &&
+		!noSample.Jackpot && noSample.NextState.PendingSpendJackpotTokens == 1 &&
+		!crossed.Jackpot && crossed.NextState.PendingSpendJackpotTokens == 3 &&
+		next.JackpotMechanism == domain.StrategyJackpotMechanismMilestone && next.NextState.PendingSpendJackpotTokens == 2 &&
+		priority.JackpotMechanism == domain.StrategyJackpotMechanismRTPCompensation && priority.NextState.PendingSpendJackpotTokens == 1 &&
+		fallback.JackpotMechanism == domain.StrategyJackpotMechanismMilestone && fallback.NextState.PendingSpendJackpotTokens == 0 &&
+		!poolBlocked.Jackpot && poolBlocked.NextState.PendingSpendJackpotTokens == 1
+	return scenarioResult{
+		Name: name, Passed: passed,
+		Summary: "验证大盘<=边界、无样本关闭、每档累计、下一子抽生效、48h优先/接续和资金不足跨轮保留",
+		Data:    rows,
+	}
+}
+
 // simulationStore models the repository contract that the pure kernel expects:
 // command idempotency and state locking wrap the draw in one critical section. The
 // simulator does not claim an in-memory mutex is the production implementation; it
diff --git a/services/lucky-gift-service/deploy/mysql/initdb/001_lucky_gift_service.sql b/services/lucky-gift-service/deploy/mysql/initdb/001_lucky_gift_service.sql
index 9a9fc2fe..026188af 100644
--- a/services/lucky-gift-service/deploy/mysql/initdb/001_lucky_gift_service.sql
+++ b/services/lucky-gift-service/deploy/mysql/initdb/001_lucky_gift_service.sql
@@ -280,7 +280,8 @@ CREATE TABLE IF NOT EXISTS lucky_user_states (
   cumulative_wager_coins BIGINT NOT NULL DEFAULT 0 COMMENT '该用户在当前奖池累计消耗金币',
   equivalent_draws BIGINT NOT NULL DEFAULT 0 COMMENT '累计金币按规则参考价格折算后的等价抽数',
   loss_streak BIGINT NOT NULL DEFAULT 0 COMMENT '连续未获得可见奖励次数，中奖后清零',
-  pending_jackpot_tokens BIGINT NOT NULL DEFAULT 0 COMMENT '未消费大奖机会；按UTC日消费产生，资金不足时允许跨日保留',
+  pending_jackpot_tokens BIGINT NOT NULL DEFAULT 0 COMMENT '旧版消费资格；新版运行时仅惰性清零，永不参与开奖',
+  pending_spend_jackpot_tokens BIGINT NOT NULL DEFAULT 0 COMMENT '新版累计消费大奖资格；受阻时跨轮次和UTC日保留，成功开奖仅扣1',
   created_at_ms BIGINT NOT NULL COMMENT '创建时间，UTC epoch ms',
   updated_at_ms BIGINT NOT NULL COMMENT '更新时间，UTC epoch ms',
   PRIMARY KEY (app_code, user_id, gift_id)
diff --git a/services/lucky-gift-service/deploy/mysql/migrations/007_dual_jackpot_spend_tokens.sql b/services/lucky-gift-service/deploy/mysql/migrations/007_dual_jackpot_spend_tokens.sql
new file mode 100644
index 00000000..2cd03f85
--- /dev/null
+++ b/services/lucky-gift-service/deploy/mysql/migrations/007_dual_jackpot_spend_tokens.sql
@@ -0,0 +1,17 @@
+SET NAMES utf8mb4 COLLATE utf8mb4_unicode_ci;
+
+USE hyapp_lucky_gift;
+
+-- 性能边界：lucky_user_states 是热状态表，本迁移只添加一个常量默认列并强制
+-- ALGORITHM=INSTANT。若目标 MySQL 不支持即时 DDL 就直接失败，绝不退化成 COPY；
+-- 上线前仍需检查长事务，避免短暂 MDL 等待。迁移不扫描、不更新历史用户行。
+-- 历史 pending_jackpot_tokens 保持原值，业务在该用户下一次付费抽奖事务内按主键惰性清零，
+-- 且从不把旧值复制到新列，因此停用期间的资格和消费都不会被追溯补发。
+SET @ddl := IF(
+  (SELECT COUNT(*) FROM information_schema.COLUMNS
+   WHERE TABLE_SCHEMA = DATABASE() AND TABLE_NAME = 'lucky_user_states'
+     AND COLUMN_NAME = 'pending_spend_jackpot_tokens') = 0,
+  'ALTER TABLE lucky_user_states ADD COLUMN pending_spend_jackpot_tokens BIGINT NOT NULL DEFAULT 0 COMMENT ''新版累计消费大奖资格；受阻时跨轮次和UTC日保留，成功开奖仅扣1'' AFTER pending_jackpot_tokens, ALGORITHM=INSTANT',
+  'SELECT 1'
+);
+PREPARE stmt FROM @ddl; EXECUTE stmt; DEALLOCATE PREPARE stmt;
diff --git a/services/lucky-gift-service/internal/domain/luckygift/strategy.go b/services/lucky-gift-service/internal/domain/luckygift/strategy.go
index ab5104e7..0cd4ffd6 100644
--- a/services/lucky-gift-service/internal/domain/luckygift/strategy.go
+++ b/services/lucky-gift-service/internal/domain/luckygift/strategy.go
@@ -26,7 +26,8 @@ const (
 	StrategyReasonPoolInsufficient          = "pool_insufficient_w_gt_p"
 	StrategyReasonRiskCapacity              = "risk_capacity_exceeded"
 	StrategyReasonNoPayableJackpot          = "no_payable_jackpot_candidate"
-	StrategyJackpotMechanismMilestone       = "milestone_token"
+	StrategyReasonSpendGlobalRTPBlocked     = "spend_global_rtp_gate_failed"
+	StrategyJackpotMechanismMilestone       = "spend_milestone"
 	StrategyJackpotMechanismRTPCompensation = "rtp_compensation"
 	StrategyDrawKindOriginal                = "original"
 	StrategyDrawKindRedraw                  = "redraw"
@@ -35,6 +36,8 @@ const (
 	StrategyRemovalMissProtection           = "miss_protection_excludes_zero"
 	StrategyRemovalPWRedrawZero             = "w_gt_p_redraw_excludes_zero"
 	StrategyConditionGlobalRTP              = "global_rtp"
+	StrategyConditionSpendGlobalRTP         = "spend_global_rtp"
+	StrategyConditionSpendDailyJackpotLimit = "spend_daily_jackpot_limit"
 	StrategyConditionSettledRound           = "settled_round_qualification"
 	StrategyConditionUserRoundRTP           = "user_round_rtp"
 	StrategyConditionUser48HourRTP          = "user_48h_rtp"
@@ -152,11 +155,18 @@ type StrategyRTP struct {
 // must update it atomically with the draw record; Version is provided for CAS/optimistic
 // locking but the kernel itself intentionally performs no I/O.
 type StrategyState struct {
-	PoolBalanceCoins        int64       `json:"pool_balance_coins"`
-	ConsecutiveZeroDraws    int64       `json:"consecutive_zero_draws"`
-	PendingMilestoneTokens  int64       `json:"pending_milestone_tokens"`
-	UserDailyJackpotWins    int64       `json:"user_daily_jackpot_wins"`
-	UserDaySpendCoins       int64       `json:"user_day_spend_coins"`
+	PoolBalanceCoins     int64 `json:"pool_balance_coins"`
+	ConsecutiveZeroDraws int64 `json:"consecutive_zero_draws"`
+	// PendingMilestoneTokens is the legacy pure-kernel field. Production dynamic_v3 never reads it:
+	// the separate field below prevents tokens created by the retired implementation from becoming
+	// a jackpot promise after the dual-strategy rollout.
+	PendingMilestoneTokens    int64 `json:"pending_milestone_tokens"`
+	PendingSpendJackpotTokens int64 `json:"pending_spend_jackpot_tokens"`
+	UserDailyJackpotWins      int64 `json:"user_daily_jackpot_wins"`
+	UserDaySpendCoins         int64 `json:"user_day_spend_coins"`
+	// SpendGlobalRTP is the latest fully closed platform window observed before this subdraw. It is
+	// deliberately independent from GlobalRTP, which is frozen with a user's one-shot 48h qualification.
+	SpendGlobalRTP          StrategyRTP `json:"spend_global_rtp"`
 	GlobalRTP               StrategyRTP `json:"global_rtp"`
 	UserRoundRTP            StrategyRTP `json:"user_round_rtp"`
 	User48HourRTP           StrategyRTP `json:"user_48h_rtp"`
@@ -255,9 +265,14 @@ type DecisionTrace struct {
 	FinalReason                string                   `json:"final_reason"`
 	BlockedReason              string                   `json:"blocked_reason,omitempty"`
 	JackpotMechanism           string                   `json:"jackpot_mechanism,omitempty"`
-	MilestoneTokenConsumed     bool                     `json:"milestone_token_consumed"`
-	MilestoneTokenRetained     bool                     `json:"milestone_token_retained"`
-	MilestoneTokensEarned      int64                    `json:"milestone_tokens_earned"`
+	MilestoneTokenConsumed     bool                     `json:"spend_milestone_token_consumed"`
+	MilestoneTokenRetained     bool                     `json:"spend_milestone_token_retained"`
+	MilestoneTokensEarned      int64                    `json:"spend_milestone_tokens_earned"`
+	// The exact before/after balance is persisted for every draw, including a successful 48h
+	// compensation that deliberately leaves an existing spend token untouched. The boolean fields
+	// above keep their event meaning; these counters make multi-token incident replay unambiguous.
+	SpendMilestoneTokensBefore int64 `json:"pending_spend_jackpot_tokens_before"`
+	SpendMilestoneTokensAfter  int64 `json:"pending_spend_jackpot_tokens_after"`
 }
 
 type StrategyDecision struct {
@@ -401,10 +416,10 @@ func PreviewLuckyGiftStrategyWeights(config StrategyConfig, state StrategyState,
 	return stage, traces, err
 }
 
-// DecideLuckyGiftStrategy is the shared production/simulation kernel. Its ordering is
-// intentional and traceable: persisted mechanism-2 token -> mechanism-1 RTP gates ->
-// ordinary weighted P/W draw and redraw -> state transition. A blocked special
-// mechanism falls through to an ordinary draw while preserving its token.
+// DecideLuckyGiftStrategy is the shared production/simulation kernel. Production V3 evaluates the
+// one-shot closed-round/48h qualification first, then the independent durable spend milestone, and
+// finally the ordinary weighted draw. A spend token is consumed only by a paid spend jackpot; all
+// failed global-RTP, daily-limit, pool and risk checks leave it available for a later paid draw.
 func DecideLuckyGiftStrategy(config StrategyConfig, state StrategyState, input StrategyInput, random StrategyRandomSource) (StrategyDecision, error) {
 	if random == nil {
 		return StrategyDecision{}, ErrStrategyRandomSource
@@ -425,48 +440,41 @@ func DecideLuckyGiftStrategy(config StrategyConfig, state StrategyState, input S
 		Stage: stage, PoolBeforeCoins: state.PoolBalanceCoins,
 		PoolContributionCoins: input.PoolContributionCoins, AvailablePoolCoins: available,
 		EffectiveRiskCapacityCoins: input.RiskCapacity.capacity(), Weights: weightTrace,
+		SpendMilestoneTokensBefore: pendingSpendJackpotTokens(config, state),
+		SpendMilestoneTokensAfter:  pendingSpendJackpotTokens(config, state),
 	}
 
-	// A persisted milestone token is a promise about the next draw, so it outranks the
-	// opportunistic RTP compensation mechanism. Insufficient pool/risk capacity keeps
-	// the token durable and lets the user retain the ordinary draw they paid for.
-	if config.JackpotMechanism2Enabled && state.PendingMilestoneTokens > 0 {
-		tier, payout, ok, reason, removed, err := selectPayableJackpot(config, state, input, available, random)
+	if config.SettledRoundRTPEligibility {
+		decision, decided, err := tryRTPCompensationJackpot(config, state, input, available, random, &trace)
 		if err != nil {
 			return StrategyDecision{}, err
 		}
-		trace.Removed = append(trace.Removed, removed...)
-		if ok {
-			trace.JackpotMechanism = StrategyJackpotMechanismMilestone
-			trace.MilestoneTokenConsumed = true
-			trace.FinalReason = StrategyReasonMilestoneJackpot
-			trace.Draws = append(trace.Draws, StrategyDrawTrace{Sequence: 1, Kind: StrategyDrawKindMilestone, TierID: tier.ID, MultiplierPPM: tier.MultiplierPPM, WeightPPM: jackpotWeight(tier), PayoutCoins: payout})
-			return finalizeStrategyDecision(config, state, input, available, tier, payout, StrategyJackpotMechanismMilestone, true, trace)
+		if decided {
+			return decision, nil
 		}
-		trace.MilestoneTokenRetained = true
-		trace.BlockedReason = reason
-		trace.Conditions = append(trace.Conditions, StrategyConditionTrace{Name: StrategyConditionPayableJackpot, Passed: false, Reason: reason})
-	}
-
-	if config.JackpotMechanism1Enabled {
-		conditions, eligible := mechanismOneConditions(config, state)
-		trace.Conditions = append(trace.Conditions, conditions...)
-		if eligible {
-			tier, payout, ok, reason, removed, err := selectPayableJackpot(config, state, input, available, random)
-			if err != nil {
-				return StrategyDecision{}, err
-			}
-			trace.Removed = append(trace.Removed, removed...)
-			trace.Conditions = append(trace.Conditions, StrategyConditionTrace{Name: StrategyConditionPayableJackpot, Passed: ok, Reason: reason})
-			if ok {
-				trace.JackpotMechanism = StrategyJackpotMechanismRTPCompensation
-				trace.FinalReason = StrategyReasonRTPCompensationJackpot
-				trace.Draws = append(trace.Draws, StrategyDrawTrace{Sequence: 1, Kind: StrategyDrawKindRTPCompensation, TierID: tier.ID, MultiplierPPM: tier.MultiplierPPM, WeightPPM: jackpotWeight(tier), PayoutCoins: payout})
-				return finalizeStrategyDecision(config, state, input, available, tier, payout, StrategyJackpotMechanismRTPCompensation, false, trace)
-			}
-			if trace.BlockedReason == "" {
-				trace.BlockedReason = reason
-			}
+		decision, decided, err = trySpendMilestoneJackpot(config, state, input, available, random, true, &trace)
+		if err != nil {
+			return StrategyDecision{}, err
+		}
+		if decided {
+			return decision, nil
+		}
+	} else {
+		// The isolated legacy kernel keeps its historical order and gate shape. Runtime dynamic_v3
+		// always takes the branch above and therefore never reads PendingMilestoneTokens.
+		decision, decided, err := trySpendMilestoneJackpot(config, state, input, available, random, false, &trace)
+		if err != nil {
+			return StrategyDecision{}, err
+		}
+		if decided {
+			return decision, nil
+		}
+		decision, decided, err = tryRTPCompensationJackpot(config, state, input, available, random, &trace)
+		if err != nil {
+			return StrategyDecision{}, err
+		}
+		if decided {
+			return decision, nil
 		}
 	}
 
@@ -485,11 +493,91 @@ func DecideLuckyGiftStrategy(config StrategyConfig, state StrategyState, input S
 	return finalizeStrategyDecision(config, state, input, available, tier, payout, "", false, trace)
 }
 
+func tryRTPCompensationJackpot(config StrategyConfig, state StrategyState, input StrategyInput, available int64, random StrategyRandomSource, trace *DecisionTrace) (StrategyDecision, bool, error) {
+	if !config.JackpotMechanism1Enabled {
+		return StrategyDecision{}, false, nil
+	}
+	conditions, eligible := mechanismOneConditions(config, state)
+	trace.Conditions = append(trace.Conditions, conditions...)
+	if !eligible {
+		return StrategyDecision{}, false, nil
+	}
+	tier, payout, ok, reason, removed, err := selectPayableJackpot(config, state, input, available, random)
+	if err != nil {
+		return StrategyDecision{}, false, err
+	}
+	trace.Removed = append(trace.Removed, removed...)
+	trace.Conditions = append(trace.Conditions, StrategyConditionTrace{Name: StrategyConditionPayableJackpot, Passed: ok, Reason: reason})
+	if !ok {
+		if trace.BlockedReason == "" {
+			trace.BlockedReason = reason
+		}
+		return StrategyDecision{}, false, nil
+	}
+	trace.JackpotMechanism = StrategyJackpotMechanismRTPCompensation
+	trace.FinalReason = StrategyReasonRTPCompensationJackpot
+	trace.Draws = append(trace.Draws, StrategyDrawTrace{Sequence: 1, Kind: StrategyDrawKindRTPCompensation, TierID: tier.ID, MultiplierPPM: tier.MultiplierPPM, WeightPPM: jackpotWeight(tier), PayoutCoins: payout})
+	decision, err := finalizeStrategyDecision(config, state, input, available, tier, payout, StrategyJackpotMechanismRTPCompensation, false, *trace)
+	return decision, err == nil, err
+}
+
+func trySpendMilestoneJackpot(config StrategyConfig, state StrategyState, input StrategyInput, available int64, random StrategyRandomSource, requireSettledGlobal bool, trace *DecisionTrace) (StrategyDecision, bool, error) {
+	if !config.JackpotMechanism2Enabled || pendingSpendJackpotTokens(config, state) <= 0 {
+		return StrategyDecision{}, false, nil
+	}
+	if requireSettledGlobal {
+		ratio, valid := state.SpendGlobalRTP.RatioPPM()
+		globalPassed := valid && rtpAtOrBelowExact(state.SpendGlobalRTP, config.GlobalRTPMaxPPM)
+		trace.Conditions = append(trace.Conditions, rtpAtOrBelowConditionTrace(StrategyConditionSpendGlobalRTP, state.SpendGlobalRTP, ratio, config.GlobalRTPMaxPPM, valid, globalPassed))
+		dailyPassed := state.UserDailyJackpotWins < config.DailyJackpotLimit
+		trace.Conditions = append(trace.Conditions, StrategyConditionTrace{
+			Name: StrategyConditionSpendDailyJackpotLimit, Numerator: state.UserDailyJackpotWins, LimitPPM: config.DailyJackpotLimit,
+			Passed: dailyPassed, Reason: boolReason(dailyPassed, "below_limit", StrategyReasonDailyJackpotLimit),
+		})
+		if !globalPassed || !dailyPassed {
+			trace.MilestoneTokenRetained = true
+			if trace.BlockedReason == "" {
+				trace.BlockedReason = StrategyReasonSpendGlobalRTPBlocked
+				if globalPassed {
+					trace.BlockedReason = StrategyReasonDailyJackpotLimit
+				}
+			}
+			return StrategyDecision{}, false, nil
+		}
+	}
+	tier, payout, ok, reason, removed, err := selectPayableJackpot(config, state, input, available, random)
+	if err != nil {
+		return StrategyDecision{}, false, err
+	}
+	trace.Removed = append(trace.Removed, removed...)
+	trace.Conditions = append(trace.Conditions, StrategyConditionTrace{Name: StrategyConditionPayableJackpot, Passed: ok, Reason: reason})
+	if !ok {
+		trace.MilestoneTokenRetained = true
+		if trace.BlockedReason == "" {
+			trace.BlockedReason = reason
+		}
+		return StrategyDecision{}, false, nil
+	}
+	trace.JackpotMechanism = StrategyJackpotMechanismMilestone
+	trace.MilestoneTokenConsumed = true
+	trace.FinalReason = StrategyReasonMilestoneJackpot
+	trace.Draws = append(trace.Draws, StrategyDrawTrace{Sequence: 1, Kind: StrategyDrawKindMilestone, TierID: tier.ID, MultiplierPPM: tier.MultiplierPPM, WeightPPM: jackpotWeight(tier), PayoutCoins: payout})
+	decision, err := finalizeStrategyDecision(config, state, input, available, tier, payout, StrategyJackpotMechanismMilestone, true, *trace)
+	return decision, err == nil, err
+}
+
+func pendingSpendJackpotTokens(config StrategyConfig, state StrategyState) int64 {
+	if config.SettledRoundRTPEligibility {
+		return state.PendingSpendJackpotTokens
+	}
+	return state.PendingMilestoneTokens
+}
+
 func validateStrategy(config StrategyConfig, state StrategyState, input StrategyInput, requireTiers bool) error {
-	if state.PoolBalanceCoins < 0 || state.ConsecutiveZeroDraws < 0 || state.PendingMilestoneTokens < 0 || state.UserDailyJackpotWins < 0 || state.UserDaySpendCoins < 0 || state.Version < 0 {
+	if state.PoolBalanceCoins < 0 || state.ConsecutiveZeroDraws < 0 || state.PendingMilestoneTokens < 0 || state.PendingSpendJackpotTokens < 0 || state.UserDailyJackpotWins < 0 || state.UserDaySpendCoins < 0 || state.Version < 0 {
 		return fmt.Errorf("%w: state counters and money cannot be negative", ErrStrategyInput)
 	}
-	for _, rtp := range []StrategyRTP{state.GlobalRTP, state.UserRoundRTP, state.User48HourRTP, state.UserDayRTP, state.User72HourRTP} {
+	for _, rtp := range []StrategyRTP{state.SpendGlobalRTP, state.GlobalRTP, state.UserRoundRTP, state.User48HourRTP, state.UserDayRTP, state.User72HourRTP} {
 		if rtp.WagerCoins < 0 || rtp.PayoutCoins < 0 {
 			return fmt.Errorf("%w: RTP numerator and denominator cannot be negative", ErrStrategyInput)
 		}
@@ -748,6 +836,16 @@ func rtpConditionTrace(name string, rtp StrategyRTP, ratio, limit int64, valid,
 	return StrategyConditionTrace{Name: name, Numerator: rtp.PayoutCoins, Denominator: rtp.WagerCoins, RatioPPM: ratio, LimitPPM: limit, Passed: passed, Reason: reason}
 }
 
+func rtpAtOrBelowConditionTrace(name string, rtp StrategyRTP, ratio, limit int64, valid, passed bool) StrategyConditionTrace {
+	reason := "within_limit"
+	if !valid {
+		reason = "denominator_zero"
+	} else if !passed {
+		reason = "above_limit"
+	}
+	return StrategyConditionTrace{Name: name, Numerator: rtp.PayoutCoins, Denominator: rtp.WagerCoins, RatioPPM: ratio, LimitPPM: limit, Passed: passed, Reason: reason}
+}
+
 func rtpBelowExact(rtp StrategyRTP, limitPPM int64) bool {
 	return compareRTPExact(rtp, limitPPM) < 0
 }
@@ -908,10 +1006,14 @@ func finalizeStrategyDecision(config StrategyConfig, state StrategyState, input
 		next.ConsecutiveZeroDraws = 0
 	}
 	if consumeToken {
-		if next.PendingMilestoneTokens <= 0 {
+		if pendingSpendJackpotTokens(config, next) <= 0 {
 			return StrategyDecision{}, fmt.Errorf("%w: cannot consume absent milestone token", ErrStrategyInput)
 		}
-		next.PendingMilestoneTokens--
+		if config.SettledRoundRTPEligibility {
+			next.PendingSpendJackpotTokens--
+		} else {
+			next.PendingMilestoneTokens--
+		}
 	}
 	jackpot := payout > 0 && tier.Jackpot
 	if jackpot {
@@ -922,21 +1024,28 @@ func finalizeStrategyDecision(config StrategyConfig, state StrategyState, input
 	if next.UserDaySpendCoins, err = safeAdd(next.UserDaySpendCoins, input.GiftPriceCoins); err != nil {
 		return StrategyDecision{}, err
 	}
-	// A threshold crossed by this draw is persisted after selection, so the newly
-	// earned token can affect only the next locked draw and never the current one. Production V3's
-	// settled-round mode has no independent spend-token qualification, so it must neither earn nor
-	// carry a hidden third jackpot condition even if an old rule row still contains the legacy value.
+	// A threshold crossed by this draw is persisted after selection, so the newly earned token can
+	// affect only the next locked subdraw and never the current one. Production V3 stores it in the
+	// dedicated field; the retired legacy token is neither read nor migrated into the new strategy.
 	earnedTokens := int64(0)
-	if !config.SettledRoundRTPEligibility {
+	if config.JackpotMechanism2Enabled {
 		earnedTokens = MilestoneTokensEarned(previousDaySpend, next.UserDaySpendCoins, config.MilestoneSpendCoins)
 	}
 	if earnedTokens > 0 {
-		next.PendingMilestoneTokens, err = safeAdd(next.PendingMilestoneTokens, earnedTokens)
+		if config.SettledRoundRTPEligibility {
+			next.PendingSpendJackpotTokens, err = safeAdd(next.PendingSpendJackpotTokens, earnedTokens)
+		} else {
+			next.PendingMilestoneTokens, err = safeAdd(next.PendingMilestoneTokens, earnedTokens)
+		}
 		if err != nil {
 			return StrategyDecision{}, err
 		}
 		trace.MilestoneTokensEarned = earnedTokens
 	}
+	// Set the final balance after both the optional consumption and threshold-crossing accrual.
+	// This assignment runs for ordinary, 48h and spend-jackpot decisions alike, so a 48h win with
+	// one pending spend promise is durably recorded as 1 -> 1 instead of looking like no promise existed.
+	trace.SpendMilestoneTokensAfter = pendingSpendJackpotTokens(config, next)
 	if !config.SettledRoundRTPEligibility {
 		if next.GlobalRTP, err = next.GlobalRTP.add(input.GiftPriceCoins, payout); err != nil {
 			return StrategyDecision{}, err
diff --git a/services/lucky-gift-service/internal/domain/luckygift/strategy_test.go b/services/lucky-gift-service/internal/domain/luckygift/strategy_test.go
index 39c79c0d..cb098bdb 100644
--- a/services/lucky-gift-service/internal/domain/luckygift/strategy_test.go
+++ b/services/lucky-gift-service/internal/domain/luckygift/strategy_test.go
@@ -294,6 +294,179 @@ func TestLuckyGiftStrategyMilestoneTokenRetainedThenConsumed(t *testing.T) {
 	}
 }
 
+func TestLuckyGiftStrategyDualJackpotSpendGlobalGateAndDurability(t *testing.T) {
+	config := strategyTestConfig()
+	config.SettledRoundRTPEligibility = true
+	config.JackpotMechanism1Enabled = true
+	config.JackpotMechanism2Enabled = true
+	config.MilestoneSpendCoins = 50
+	config.GlobalRTPMaxPPM = 980_000
+	config.UserRoundRTPMaxPPM = 960_000
+	config.User48HourRTPMaxPPM = 960_000
+	config.DailyJackpotLimit = 5
+
+	base := StrategyState{PoolBalanceCoins: 10_000, PendingSpendJackpotTokens: 1, SpendGlobalRTP: StrategyRTP{WagerCoins: 100, PayoutCoins: 98}}
+	paid, err := DecideLuckyGiftStrategy(config, base, strategyTestInput(10), NewSeededStrategyRandom(80))
+	if err != nil {
+		t.Fatal(err)
+	}
+	if paid.JackpotMechanism != StrategyJackpotMechanismMilestone || !paid.ConsumedMilestoneToken || paid.NextState.PendingSpendJackpotTokens != 0 ||
+		paid.Trace.SpendMilestoneTokensBefore != 1 || paid.Trace.SpendMilestoneTokensAfter != 0 {
+		t.Fatalf("global equality should pay exactly one spend token: %+v", paid)
+	}
+
+	for _, test := range []struct {
+		name string
+		rtp  StrategyRTP
+	}{
+		{name: "no closed platform sample", rtp: StrategyRTP{}},
+		{name: "platform above configured maximum", rtp: StrategyRTP{WagerCoins: 100, PayoutCoins: 99}},
+	} {
+		t.Run(test.name, func(t *testing.T) {
+			state := base
+			state.SpendGlobalRTP = test.rtp
+			blocked, err := DecideLuckyGiftStrategy(config, state, strategyTestInput(10), NewSeededStrategyRandom(81))
+			if err != nil {
+				t.Fatal(err)
+			}
+			if blocked.JackpotMechanism != "" || blocked.NextState.PendingSpendJackpotTokens != 1 || !blocked.Trace.MilestoneTokenRetained {
+				t.Fatalf("blocked spend token must remain durable: %+v", blocked)
+			}
+		})
+	}
+
+	capState := base
+	capState.UserDailyJackpotWins = config.DailyJackpotLimit
+	blockedBySharedCap, err := DecideLuckyGiftStrategy(config, capState, strategyTestInput(10), NewSeededStrategyRandom(82))
+	if err != nil {
+		t.Fatal(err)
+	}
+	if blockedBySharedCap.NextState.PendingSpendJackpotTokens != 1 || blockedBySharedCap.Jackpot {
+		t.Fatalf("daily cap must be shared without consuming spend token: %+v", blockedBySharedCap)
+	}
+}
+
+func TestLuckyGiftStrategyDualJackpotCrossingAndPriority(t *testing.T) {
+	config := strategyTestConfig()
+	config.SettledRoundRTPEligibility = true
+	config.JackpotMechanism1Enabled = true
+	config.JackpotMechanism2Enabled = true
+	config.MilestoneSpendCoins = 50
+	config.GlobalRTPMaxPPM = 980_000
+	config.UserRoundRTPMaxPPM = 960_000
+	config.User48HourRTPMaxPPM = 960_000
+	config.DailyJackpotLimit = 5
+
+	// One paid subdraw can cross several exact multiples, but its new promises are appended only
+	// after selection. Even with a valid global sample this draw must stay on the ordinary path.
+	crossingState := StrategyState{
+		PoolBalanceCoins: 10_000, UserDaySpendCoins: 40,
+		SpendGlobalRTP: StrategyRTP{WagerCoins: 100, PayoutCoins: 98},
+	}
+	crossing, err := DecideLuckyGiftStrategy(config, crossingState, strategyTestInput(110), &scriptedStrategyRandom{t: t, indexes: []int64{0}, bounds: []int64{1_000_000}})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if crossing.Jackpot || crossing.NextState.PendingSpendJackpotTokens != 3 || crossing.Trace.MilestoneTokensEarned != 3 ||
+		crossing.Trace.SpendMilestoneTokensBefore != 0 || crossing.Trace.SpendMilestoneTokensAfter != 3 {
+		t.Fatalf("current draw must only earn three next-draw tokens: %+v", crossing)
+	}
+
+	const hourMS = int64(60 * 60 * 1000)
+	qualified := StrategyState{
+		PoolBalanceCoins:          10_000,
+		PendingSpendJackpotTokens: 1,
+		SpendGlobalRTP:            StrategyRTP{WagerCoins: 100, PayoutCoins: 98},
+		SettledRoundPending:       true,
+		GlobalRTP:                 StrategyRTP{WagerCoins: 100, PayoutCoins: 98},
+		UserRoundRTP:              StrategyRTP{WagerCoins: 100, PayoutCoins: 95},
+		User48HourRTP:             StrategyRTP{WagerCoins: 10_000, PayoutCoins: 9_500},
+		UserRegisteredAtMS:        hourMS,
+		QualificationClosedAtMS:   49 * hourMS,
+	}
+	compensated, err := DecideLuckyGiftStrategy(config, qualified, strategyTestInput(10), NewSeededStrategyRandom(83))
+	if err != nil {
+		t.Fatal(err)
+	}
+	if compensated.JackpotMechanism != StrategyJackpotMechanismRTPCompensation || compensated.NextState.PendingSpendJackpotTokens != 1 || compensated.NextState.SettledRoundPending ||
+		compensated.Trace.SpendMilestoneTokensBefore != 1 || compensated.Trace.SpendMilestoneTokensAfter != 1 {
+		t.Fatalf("48h qualification must win first without consuming spend token: %+v", compensated)
+	}
+
+	// A failed personal RTP condition consumes only the one-shot 48h qualification. The same child
+	// draw may still use the independent spend promise because registration and personal RTP do not gate it.
+	qualified.UserRoundRTP.PayoutCoins = 96
+	spendFallback, err := DecideLuckyGiftStrategy(config, qualified, strategyTestInput(10), NewSeededStrategyRandom(84))
+	if err != nil {
+		t.Fatal(err)
+	}
+	if spendFallback.JackpotMechanism != StrategyJackpotMechanismMilestone || spendFallback.NextState.PendingSpendJackpotTokens != 0 || spendFallback.NextState.SettledRoundPending {
+		t.Fatalf("ineligible 48h path should fall through to spend jackpot: %+v", spendFallback)
+	}
+
+	qualified.PoolBalanceCoins = 49
+	qualified.UserRoundRTP.PayoutCoins = 95
+	poolBlocked, err := DecideLuckyGiftStrategy(config, qualified, strategyTestInput(10), NewSeededStrategyRandom(85))
+	if err != nil {
+		t.Fatal(err)
+	}
+	if poolBlocked.Jackpot || poolBlocked.NextState.SettledRoundPending || poolBlocked.NextState.PendingSpendJackpotTokens != 1 {
+		t.Fatalf("48h is one-shot while pool-blocked spend token remains: %+v", poolBlocked)
+	}
+}
+
+func TestLuckyGiftStrategySpendJackpotSharesAllSixInclusiveRiskCaps(t *testing.T) {
+	config := strategyTestConfig()
+	config.SettledRoundRTPEligibility = true
+	config.JackpotMechanism1Enabled = true
+	config.JackpotMechanism2Enabled = true
+	config.MilestoneSpendCoins = 50
+	config.GlobalRTPMaxPPM = 980_000
+	config.UserRoundRTPMaxPPM = 960_000
+	config.User48HourRTPMaxPPM = 960_000
+	config.DailyJackpotLimit = 5
+	setters := []struct {
+		name string
+		set  func(*StrategyRiskCapacity, int64)
+	}{
+		{name: "single", set: func(c *StrategyRiskCapacity, v int64) { c.SingleDrawCoins = v }},
+		{name: "user-hour", set: func(c *StrategyRiskCapacity, v int64) { c.UserHourCoins = v }},
+		{name: "user-day", set: func(c *StrategyRiskCapacity, v int64) { c.UserDayCoins = v }},
+		{name: "device-day", set: func(c *StrategyRiskCapacity, v int64) { c.DeviceDayCoins = v }},
+		{name: "room-hour", set: func(c *StrategyRiskCapacity, v int64) { c.RoomHourCoins = v }},
+		{name: "anchor-day", set: func(c *StrategyRiskCapacity, v int64) { c.AnchorDayCoins = v }},
+	}
+	for _, test := range setters {
+		t.Run(test.name, func(t *testing.T) {
+			state := StrategyState{
+				PoolBalanceCoins: 10_000, PendingSpendJackpotTokens: 1,
+				SpendGlobalRTP: StrategyRTP{WagerCoins: 100, PayoutCoins: 98},
+			}
+			capacity := StrategyRiskCapacity{Enabled: true, SingleDrawCoins: 10_000, UserHourCoins: 10_000, UserDayCoins: 10_000, DeviceDayCoins: 10_000, RoomHourCoins: 10_000, AnchorDayCoins: 10_000}
+			test.set(&capacity, 1_999)
+			input := strategyTestInput(10)
+			input.RiskCapacity = capacity
+			blocked, err := DecideLuckyGiftStrategy(config, state, input, NewSeededStrategyRandom(90))
+			if err != nil {
+				t.Fatal(err)
+			}
+			if blocked.Jackpot || blocked.NextState.PendingSpendJackpotTokens != 1 {
+				t.Fatalf("capacity below minimum jackpot must retain token: %+v", blocked)
+			}
+
+			test.set(&capacity, 2_000)
+			input.RiskCapacity = capacity
+			paid, err := DecideLuckyGiftStrategy(config, state, input, NewSeededStrategyRandom(91))
+			if err != nil {
+				t.Fatal(err)
+			}
+			if !paid.Jackpot || paid.PayoutCoins != 2_000 || paid.NextState.PendingSpendJackpotTokens != 0 {
+				t.Fatalf("capacity equality must permit the 200x jackpot: %+v", paid)
+			}
+		})
+	}
+}
+
 func TestLuckyGiftStrategyDailyJackpotLimitDoesNotConstrainOrdinarySameMultiplier(t *testing.T) {
 	config := strategyTestConfig()
 	config.JackpotMechanism2Enabled = true
diff --git a/services/lucky-gift-service/internal/service/luckygift/config.go b/services/lucky-gift-service/internal/service/luckygift/config.go
index 1152ad7e..0acbc24d 100644
--- a/services/lucky-gift-service/internal/service/luckygift/config.go
+++ b/services/lucky-gift-service/internal/service/luckygift/config.go
@@ -244,9 +244,9 @@ func validateDynamicRuleConfig(config domain.RuleConfig, stages map[string]domai
 		// disabled 草稿允许金额口径仍为 0；启用前必须由各 App 按自身业务口径和风控预算显式补齐。
 		return nil
 	}
-	if !config.SettledRoundRTPEligibility && config.JackpotSpendThresholdCoins <= 0 {
-		// Only the old draft contract required the now-retired milestone amount. New round/48h configs
-		// preserve the column for read compatibility but never require or consume it.
+	if config.JackpotSpendThresholdCoins <= 0 {
+		// Enabled V3 runs the 48h compensation and spend-milestone strategies together. A zero amount
+		// would silently disable one of the two published promises, so only disabled drafts may keep 0.
 		return xerr.New(xerr.InvalidArgument, "dynamic_v3 jackpot cumulative spend threshold coins must be configured as a positive amount before enabling")
 	}
 	return validateDynamicPayoutCaps(config, true)
diff --git a/services/lucky-gift-service/internal/storage/mysql/lucky_gift_dynamic_repository.go b/services/lucky-gift-service/internal/storage/mysql/lucky_gift_dynamic_repository.go
index 8b04352c..5853de09 100644
--- a/services/lucky-gift-service/internal/storage/mysql/lucky_gift_dynamic_repository.go
+++ b/services/lucky-gift-service/internal/storage/mysql/lucky_gift_dynamic_repository.go
@@ -94,6 +94,12 @@ func (r *Repository) executeDynamicLuckyGiftDrawBatch(ctx context.Context, tx *s
 	if err != nil {
 		return nil, err
 	}
+	// Consumption milestones use the newest fully closed platform sample, independent of whether
+	// this user participated in that round. Missing history deliberately leaves a zero denominator.
+	spendGlobalWindow, hasSpendGlobalWindow, err := r.getLatestClosedLuckyDynamicRTPWindow(ctx, tx, appCode, windowScopeID, globalWindow.WindowIndex)
+	if err != nil {
+		return nil, err
+	}
 	giftWindow, err := r.getOpenLuckyDynamicRTPWindow(ctx, tx, appCode, "pool_gift", windowScopeID, config.SettlementWindowWager, config.GiftWindowDraws, config.TargetRTPPPM, paidAtMS, nowMS)
 	if err != nil {
 		return nil, err
@@ -151,16 +157,19 @@ func (r *Repository) executeDynamicLuckyGiftDrawBatch(ctx context.Context, tx *s
 		Strategy: domain.StrategyState{
 			PoolBalanceCoins:     basePool.Balance,
 			ConsecutiveZeroDraws: userState.LossStreak,
-			// Consumption milestones were retired from dynamic_v3 jackpot eligibility. Old durable tokens
-			// are intentionally cleared on the next paid draw rather than becoming a hidden third gate.
-			PendingMilestoneTokens: 0,
-			UserDailyJackpotWins:   dayState.JackpotHits,
-			UserDaySpendCoins:      dayState.SpendCoins,
-			UserRegisteredAtMS:     cmd.UserRegisteredAtMS,
-			UserDayRTP:             domain.StrategyRTP{WagerCoins: dayState.WagerCoins, PayoutCoins: dayState.PayoutCoins},
-			Version:                userState.PaidDraws,
+			// The old pending_jackpot_tokens column is intentionally not copied. It is cleared in the same
+			// transaction while only the new column can authorize the restored spend strategy.
+			PendingSpendJackpotTokens: userState.PendingSpendJackpotTokens,
+			UserDailyJackpotWins:      dayState.JackpotHits,
+			UserDaySpendCoins:         dayState.SpendCoins,
+			UserRegisteredAtMS:        cmd.UserRegisteredAtMS,
+			UserDayRTP:                domain.StrategyRTP{WagerCoins: dayState.WagerCoins, PayoutCoins: dayState.PayoutCoins},
+			Version:                   userState.PaidDraws,
 		},
 	}
+	if hasSpendGlobalWindow {
+		state.Strategy.SpendGlobalRTP = domain.StrategyRTP{WagerCoins: spendGlobalWindow.WagerCoins, PayoutCoins: spendGlobalWindow.ActualPayoutCoins}
+	}
 	if hasQualification {
 		if err := r.prepareLuckySettledRoundQualification(ctx, tx, appCode, config, cmd, qualificationGlobal, qualificationWindow, &state, nowMS); err != nil {
 			return nil, err
@@ -241,6 +250,9 @@ func (r *Repository) rollLuckyDynamicWindows(ctx context.Context, tx *sql.Tx, ap
 			closed.Status = "closed"
 			closed.ClosedAtMS = maxInt64(closed.LastPaidAtMS, paidAtMS)
 		}
+		// The next subdraw in this batch must immediately see the just-closed platform sample. The draw
+		// that completed the window was already selected before this assignment and cannot use it.
+		state.Strategy.SpendGlobalRTP = domain.StrategyRTP{WagerCoins: closed.WagerCoins, PayoutCoins: closed.ActualPayoutCoins}
 		closedUser := state.UserWindow
 		if err := r.prepareLuckySettledRoundQualification(ctx, tx, appCode, config, cmd, closed, closedUser, state, nowMS); err != nil {
 			return err
@@ -365,7 +377,8 @@ func (r *Repository) stageLuckyDynamicDecision(appCode string, config domain.Con
 	state.UserState.CumulativeWagerCoins += cmd.CoinSpent
 	state.UserState.EquivalentDraws = luckyEquivalentDraws(state.UserState.CumulativeWagerCoins, referencePrice)
 	state.UserState.LossStreak = decision.NextState.ConsecutiveZeroDraws
-	state.UserState.PendingJackpotTokens = decision.NextState.PendingMilestoneTokens
+	state.UserState.PendingJackpotTokens = 0
+	state.UserState.PendingSpendJackpotTokens = decision.NextState.PendingSpendJackpotTokens
 	state.DayState.WagerCoins += cmd.CoinSpent
 	state.DayState.PayoutCoins += decision.PayoutCoins
 	state.DayState.SpendCoins = decision.NextState.UserDaySpendCoins
@@ -509,10 +522,10 @@ func (r *Repository) persistLuckyDynamicBatch(ctx context.Context, tx *sql.Tx, a
 	if _, err := tx.ExecContext(ctx, `
 		UPDATE lucky_user_states
 		SET paid_draws = ?, cumulative_wager_coins = ?, equivalent_draws = ?, loss_streak = ?,
-		    pending_jackpot_tokens = ?, updated_at_ms = ?
+		    pending_jackpot_tokens = 0, pending_spend_jackpot_tokens = ?, updated_at_ms = ?
 		WHERE app_code = ? AND user_id = ? AND gift_id = ?`,
 		state.UserState.PaidDraws, state.UserState.CumulativeWagerCoins, state.UserState.EquivalentDraws, state.UserState.LossStreak,
-		state.UserState.PendingJackpotTokens, nowMS, appCode, cmd.UserID, config.GiftID,
+		state.UserState.PendingSpendJackpotTokens, nowMS, appCode, cmd.UserID, config.GiftID,
 	); err != nil {
 		return err
 	}
diff --git a/services/lucky-gift-service/internal/storage/mysql/lucky_gift_dynamic_repository_test.go b/services/lucky-gift-service/internal/storage/mysql/lucky_gift_dynamic_repository_test.go
index 560b4fa9..c9863001 100644
--- a/services/lucky-gift-service/internal/storage/mysql/lucky_gift_dynamic_repository_test.go
+++ b/services/lucky-gift-service/internal/storage/mysql/lucky_gift_dynamic_repository_test.go
@@ -131,6 +131,8 @@ func TestDynamicLuckyGiftMySQLBatchAndExternalPaths(t *testing.T) {
 	}
 	assertDynamicMySQLBatchFacts(t, schema.DB, "lalu", internalRule.PoolID, command.UserID, command.CommandID, 99, 100, 98, 1, 1)
 	assertDynamicTokenSurvivesUTCDayBoundary(t, repo, schema.DB, ctx, now)
+	assertDynamicSpendBatchRollbackAndRestart(t, repo, schema.DB, ctx, now+10_000)
+	assertExternalNewUserSpendBatch(t, repo, schema.DB, ctx, now+20_000)
 
 	externalRule := dynamicMySQLTestRule("external-dynamic")
 	if _, err := repo.PublishLuckyGiftRuleConfig(ctx, externalRule, now-500); err != nil {
@@ -244,9 +246,10 @@ func seedDynamicRTPWindowsBelowWagerThreshold(t *testing.T, db *sql.DB, poolID s
 		if _, err := db.Exec(`
 			INSERT INTO lucky_rtp_windows (
 			 app_code,scope_type,scope_id,window_index,target_rtp_ppm,control_window_draws,
-			 paid_draws,wager_coins,target_payout_coins,actual_base_payout_coins,carry_ppm,status,created_at_ms,updated_at_ms
-			) VALUES ('lalu',?,?,1,980000,10,10,98,96,0,40000,'open',?,?)`,
-			scopeType, scopeID, nowMS, nowMS,
+			 paid_draws,wager_coins,target_payout_coins,actual_base_payout_coins,carry_ppm,status,
+			 started_at_ms,last_paid_at_ms,closed_at_ms,created_at_ms,updated_at_ms
+			) VALUES ('lalu',?,?,1,980000,10,10,98,96,0,40000,'open',?,?,0,?,?)`,
+			scopeType, scopeID, nowMS-1_000, nowMS, nowMS, nowMS,
 		); err != nil {
 			t.Fatalf("seed %s RTP window: %v", scopeType, err)
 		}
@@ -316,6 +319,11 @@ func assertDynamicTokenSurvivesUTCDayBoundary(t *testing.T, repo *Repository, db
 		t.Fatalf("earn milestone token: %v", err)
 	}
 	assertPendingTokenCount(t, db, rule.PoolID, userID, 1)
+	// A retired token is deliberately planted after the new token was earned. The next transaction
+	// must clear it without copying or adding it to the new durable balance.
+	if _, err := db.Exec(`UPDATE lucky_user_states SET pending_jackpot_tokens=7 WHERE app_code='lalu' AND user_id=? AND gift_id=?`, userID, rule.PoolID); err != nil {
+		t.Fatalf("seed retired pending token: %v", err)
+	}
 	if result, err := db.Exec(`UPDATE lucky_pools SET balance=0 WHERE app_code='lalu' AND scope_type=? AND scope_id=?`, luckyDynamicPoolScopeType, rule.PoolID); err != nil {
 		t.Fatalf("drain token test pool: %v", err)
 	} else if rows, _ := result.RowsAffected(); rows != 1 {
@@ -331,11 +339,27 @@ func assertDynamicTokenSurvivesUTCDayBoundary(t *testing.T, repo *Repository, db
 		t.Fatalf("retain token across UTC day: %v", err)
 	}
 	assertPendingTokenCount(t, db, rule.PoolID, userID, 1)
+	var retired int64
+	if err := db.QueryRow(`SELECT pending_jackpot_tokens FROM lucky_user_states WHERE app_code='lalu' AND user_id=? AND gift_id=?`, userID, rule.PoolID).Scan(&retired); err != nil || retired != 0 {
+		t.Fatalf("retired token was not lazily cleared: got=%d err=%v", retired, err)
+	}
 	if result, err := db.Exec(`UPDATE lucky_pools SET balance=1000000 WHERE app_code='lalu' AND scope_type=? AND scope_id=?`, luckyDynamicPoolScopeType, rule.PoolID); err != nil {
 		t.Fatalf("refill token test pool: %v", err)
 	} else if rows, _ := result.RowsAffected(); rows != 1 {
 		t.Fatalf("refill token test pool affected=%d, want 1", rows)
 	}
+	// Freeze the current platform window as a valid <=98% sample. Until this close, the spend token
+	// correctly remains blocked because an open/current RTP can never authorize a spend jackpot.
+	if result, err := db.Exec(`
+		UPDATE lucky_rtp_windows
+		SET wager_coins=1000000, actual_base_payout_coins=980000, status='closed',
+		    started_at_ms=?, last_paid_at_ms=?, closed_at_ms=?
+		WHERE app_code='lalu' AND scope_type='pool' AND scope_id LIKE CONCAT(?, ':v%') AND status='open'`,
+		dayOneMS, dayTwoMS, dayTwoMS, rule.PoolID); err != nil {
+		t.Fatalf("close platform RTP sample: %v", err)
+	} else if rows, _ := result.RowsAffected(); rows != 1 {
+		t.Fatalf("closed platform RTP windows=%d want=1", rows)
+	}
 	consume := retain
 	consume.CommandID = "token-consume-next-day"
 	consume.PaidAtMS++
@@ -349,11 +373,228 @@ func assertDynamicTokenSurvivesUTCDayBoundary(t *testing.T, repo *Repository, db
 	assertPendingTokenCount(t, db, rule.PoolID, userID, 0)
 }
 
+// assertDynamicSpendBatchRollbackAndRestart exercises the production transaction rather than only
+// the pure kernel: child 2 earns the token after selection and child 3 consumes it. A trigger fails
+// the first attempt at the final fact insert, proving all earlier in-transaction state transitions
+// roll back; rebuilding Repository and retrying the same command must then apply the sequence once.
+func assertDynamicSpendBatchRollbackAndRestart(t *testing.T, repo *Repository, db *sql.DB, ctx context.Context, nowMS int64) {
+	t.Helper()
+	rule := dynamicMySQLTestRule("dual-spend-batch")
+	rule.JackpotSpendThresholdCoins = 20
+	published, err := repo.PublishLuckyGiftRuleConfig(ctx, rule, nowMS-2_000)
+	if err != nil {
+		t.Fatalf("publish dual spend batch rule: %v", err)
+	}
+	fundDynamicSpendTestPool(t, repo, ctx, published.PoolID, "dual-spend-batch-credit", nowMS-1_900)
+	seedClosedDynamicSpendGlobalWindow(t, db, published, nowMS-1_500)
+
+	command := domain.DrawCommand{
+		CommandID: "dual-spend-batch-command", PoolID: published.PoolID, UserID: 4_040, TargetUserID: 1,
+		DeviceID: "dual-spend-batch-device", RoomID: "dual-spend-batch-room", AnchorID: "dual-spend-batch-anchor", GiftID: "super-lucky",
+		GiftCount: 3, CoinSpent: 30, GiftIncomeCoins: 0, PaidAtMS: nowMS,
+	}
+	const rollbackTrigger = "trg_lucky_dual_spend_rollback"
+	if _, err := db.Exec("DROP TRIGGER IF EXISTS " + rollbackTrigger); err != nil {
+		t.Fatalf("drop stale dual spend rollback trigger: %v", err)
+	}
+	t.Cleanup(func() { _, _ = db.Exec("DROP TRIGGER IF EXISTS " + rollbackTrigger) })
+	if _, err := db.Exec(`
+		CREATE TRIGGER trg_lucky_dual_spend_rollback
+		BEFORE INSERT ON lucky_draw_records
+		FOR EACH ROW SIGNAL SQLSTATE '45000' SET MESSAGE_TEXT = 'forced dual spend rollback'`); err != nil {
+		t.Fatalf("create dual spend rollback trigger: %v", err)
+	}
+	if _, err := repo.ExecuteLuckyGiftDraw(ctx, command, nowMS); err == nil {
+		t.Fatal("forced persistence failure unexpectedly committed dual spend batch")
+	}
+	if _, err := db.Exec("DROP TRIGGER " + rollbackTrigger); err != nil {
+		t.Fatalf("drop dual spend rollback trigger: %v", err)
+	}
+
+	var stateRows, drawRows, dayRows, balanceAfterRollback int64
+	if err := db.QueryRow(`SELECT COUNT(*) FROM lucky_user_states WHERE app_code='lalu' AND user_id=? AND gift_id=?`, command.UserID, published.PoolID).Scan(&stateRows); err != nil {
+		t.Fatalf("query rolled-back spend state: %v", err)
+	}
+	if err := db.QueryRow(`SELECT COUNT(*) FROM lucky_draw_records WHERE app_code='lalu' AND command_id LIKE 'dual-spend-batch-command#%'`).Scan(&drawRows); err != nil {
+		t.Fatalf("query rolled-back spend draws: %v", err)
+	}
+	if err := db.QueryRow(`SELECT COUNT(*) FROM lucky_user_strategy_days WHERE app_code='lalu' AND pool_id=? AND user_id=?`, published.PoolID, command.UserID).Scan(&dayRows); err != nil {
+		t.Fatalf("query rolled-back spend day: %v", err)
+	}
+	if err := db.QueryRow(`SELECT balance FROM lucky_pools WHERE app_code='lalu' AND scope_type=? AND scope_id=?`, luckyDynamicPoolScopeType, published.PoolID).Scan(&balanceAfterRollback); err != nil {
+		t.Fatalf("query rolled-back spend pool: %v", err)
+	}
+	if stateRows != 0 || drawRows != 0 || dayRows != 0 || balanceAfterRollback != 1_000_000 {
+		t.Fatalf("failed spend batch leaked state/draw/day/pool=%d/%d/%d/%d", stateRows, drawRows, dayRows, balanceAfterRollback)
+	}
+
+	restarted := &Repository{db: db}
+	first, err := restarted.ExecuteLuckyGiftDraw(ctx, command, nowMS+1)
+	if err != nil {
+		t.Fatalf("retry dual spend batch after repository restart: %v", err)
+	}
+	assertSpendTraceSnapshot(t, db, "lucky_draw_records", "command_id", luckyDrawSubCommandID(command.CommandID, 1, 3), 0, 0, 0, "")
+	assertSpendTraceSnapshot(t, db, "lucky_draw_records", "command_id", luckyDrawSubCommandID(command.CommandID, 2, 3), 0, 1, 1, "")
+	assertSpendTraceSnapshot(t, db, "lucky_draw_records", "command_id", luckyDrawSubCommandID(command.CommandID, 3, 3), 1, 0, 0, domain.StrategyJackpotMechanismMilestone)
+
+	var paidDraws, pendingTokens, daySpend, committedDraws, balanceBeforeReplay int64
+	if err := db.QueryRow(`SELECT paid_draws, pending_spend_jackpot_tokens FROM lucky_user_states WHERE app_code='lalu' AND user_id=? AND gift_id=?`, command.UserID, published.PoolID).Scan(&paidDraws, &pendingTokens); err != nil {
+		t.Fatalf("query committed dual spend state: %v", err)
+	}
+	if err := db.QueryRow(`SELECT spend_coins FROM lucky_user_strategy_days WHERE app_code='lalu' AND pool_id=? AND user_id=? AND stat_day=?`, published.PoolID, command.UserID, luckyPaidTime(command, nowMS).Format("2006-01-02")).Scan(&daySpend); err != nil {
+		t.Fatalf("query committed dual spend day: %v", err)
+	}
+	if err := db.QueryRow(`SELECT COUNT(*) FROM lucky_draw_records WHERE app_code='lalu' AND command_id LIKE 'dual-spend-batch-command#%'`).Scan(&committedDraws); err != nil {
+		t.Fatalf("query committed dual spend draws: %v", err)
+	}
+	if err := db.QueryRow(`SELECT balance FROM lucky_pools WHERE app_code='lalu' AND scope_type=? AND scope_id=?`, luckyDynamicPoolScopeType, published.PoolID).Scan(&balanceBeforeReplay); err != nil {
+		t.Fatalf("query committed dual spend pool: %v", err)
+	}
+	if paidDraws != 3 || pendingTokens != 0 || daySpend != 30 || committedDraws != 3 {
+		t.Fatalf("committed spend batch state paid/token/day/draws=%d/%d/%d/%d", paidDraws, pendingTokens, daySpend, committedDraws)
+	}
+
+	replayed, err := (&Repository{db: db}).ExecuteLuckyGiftDraw(ctx, command, nowMS+2)
+	if err != nil {
+		t.Fatalf("replay committed dual spend batch: %v", err)
+	}
+	var replayPaidDraws, replayPendingTokens, replayDaySpend, replayDraws, balanceAfterReplay int64
+	if err := db.QueryRow(`SELECT paid_draws, pending_spend_jackpot_tokens FROM lucky_user_states WHERE app_code='lalu' AND user_id=? AND gift_id=?`, command.UserID, published.PoolID).Scan(&replayPaidDraws, &replayPendingTokens); err != nil {
+		t.Fatalf("query replayed dual spend state: %v", err)
+	}
+	if err := db.QueryRow(`SELECT spend_coins FROM lucky_user_strategy_days WHERE app_code='lalu' AND pool_id=? AND user_id=? AND stat_day=?`, published.PoolID, command.UserID, luckyPaidTime(command, nowMS).Format("2006-01-02")).Scan(&replayDaySpend); err != nil {
+		t.Fatalf("query replayed dual spend day: %v", err)
+	}
+	if err := db.QueryRow(`SELECT COUNT(*) FROM lucky_draw_records WHERE app_code='lalu' AND command_id LIKE 'dual-spend-batch-command#%'`).Scan(&replayDraws); err != nil {
+		t.Fatalf("query replayed dual spend draws: %v", err)
+	}
+	if err := db.QueryRow(`SELECT balance FROM lucky_pools WHERE app_code='lalu' AND scope_type=? AND scope_id=?`, luckyDynamicPoolScopeType, published.PoolID).Scan(&balanceAfterReplay); err != nil {
+		t.Fatalf("query replayed dual spend pool: %v", err)
+	}
+	if replayed.DrawID != first.DrawID || replayPaidDraws != 3 || replayPendingTokens != 0 || replayDaySpend != 30 || replayDraws != 3 || balanceAfterReplay != balanceBeforeReplay {
+		t.Fatalf("replay duplicated spend batch: first/replay=%s/%s paid/token/day/draws/pool=%d/%d/%d/%d/%d want pool=%d", first.DrawID, replayed.DrawID, replayPaidDraws, replayPendingTokens, replayDaySpend, replayDraws, balanceAfterReplay, balanceBeforeReplay)
+	}
+}
+
+// assertExternalNewUserSpendBatch proves the external path needs neither recharge facts nor a
+// 48-hour-old account: a one-hour-old authenticated user earns at child 2 and spends at child 3.
+func assertExternalNewUserSpendBatch(t *testing.T, repo *Repository, db *sql.DB, ctx context.Context, nowMS int64) {
+	t.Helper()
+	rule := dynamicMySQLTestRule("dual-spend-external-new-user")
+	rule.JackpotSpendThresholdCoins = 20
+	published, err := repo.PublishLuckyGiftRuleConfig(ctx, rule, nowMS-2_000)
+	if err != nil {
+		t.Fatalf("publish external new-user spend rule: %v", err)
+	}
+	fundDynamicSpendTestPool(t, repo, ctx, published.PoolID, "dual-spend-external-credit", nowMS-1_900)
+	seedClosedDynamicSpendGlobalWindow(t, db, published, nowMS-1_500)
+
+	command := domain.ExternalDrawCommand{
+		AppCode: "lalu", ExternalUserID: "dual-spend-new-external-user", DeviceID: "dual-spend-new-external-device",
+		RequestID: "dual-spend-new-external-request", GiftCount: 3, UnitAmount: 10, TotalAmount: 30,
+		Currency: "COIN", PaidAtMS: nowMS, UserRegisteredAtMS: nowMS - int64(time.Hour/time.Millisecond), PoolID: published.PoolID,
+	}
+	first, err := (&Repository{db: db}).ExecuteExternalGiftDraw(ctx, command, nowMS)
+	if err != nil {
+		t.Fatalf("execute external new-user spend batch: %v", err)
+	}
+	assertSpendTraceSnapshot(t, db, "external_lucky_gift_draw_items", "request_id", command.RequestID, 1, 0, 0, domain.StrategyJackpotMechanismMilestone)
+
+	externalUserID := externalLuckyUserID(command.AppCode, command.ExternalUserID)
+	var paidDraws, pendingTokens, itemCount, aggregateCount int64
+	if err := db.QueryRow(`SELECT paid_draws, pending_spend_jackpot_tokens FROM lucky_user_states WHERE app_code='lalu' AND user_id=? AND gift_id=?`, externalUserID, published.PoolID).Scan(&paidDraws, &pendingTokens); err != nil {
+		t.Fatalf("query external new-user spend state: %v", err)
+	}
+	if err := db.QueryRow(`SELECT COUNT(*) FROM external_lucky_gift_draw_items WHERE app_code='lalu' AND request_id=?`, command.RequestID).Scan(&itemCount); err != nil {
+		t.Fatalf("query external new-user spend items: %v", err)
+	}
+	if err := db.QueryRow(`SELECT COUNT(*) FROM external_lucky_gift_draws WHERE app_code='lalu' AND request_id=?`, command.RequestID).Scan(&aggregateCount); err != nil {
+		t.Fatalf("query external new-user spend aggregate: %v", err)
+	}
+	if first.RewardAmount <= 0 || paidDraws != 3 || pendingTokens != 0 || itemCount != 3 || aggregateCount != 1 {
+		t.Fatalf("external new-user spend result reward/paid/token/items/aggregate=%d/%d/%d/%d/%d", first.RewardAmount, paidDraws, pendingTokens, itemCount, aggregateCount)
+	}
+
+	replayed, err := (&Repository{db: db}).ExecuteExternalGiftDraw(ctx, command, nowMS+1)
+	if err != nil {
+		t.Fatalf("replay external new-user spend batch: %v", err)
+	}
+	var replayPaidDraws, replayItems int64
+	if err := db.QueryRow(`SELECT paid_draws FROM lucky_user_states WHERE app_code='lalu' AND user_id=? AND gift_id=?`, externalUserID, published.PoolID).Scan(&replayPaidDraws); err != nil {
+		t.Fatalf("query replayed external new-user state: %v", err)
+	}
+	if err := db.QueryRow(`SELECT COUNT(*) FROM external_lucky_gift_draw_items WHERE app_code='lalu' AND request_id=?`, command.RequestID).Scan(&replayItems); err != nil {
+		t.Fatalf("query replayed external new-user items: %v", err)
+	}
+	if replayed.DrawID != first.DrawID || replayPaidDraws != 3 || replayItems != 3 {
+		t.Fatalf("external retry duplicated spend batch: first/replay=%s/%s paid/items=%d/%d", first.DrawID, replayed.DrawID, replayPaidDraws, replayItems)
+	}
+}
+
+func fundDynamicSpendTestPool(t *testing.T, repo *Repository, ctx context.Context, poolID, adjustmentID string, nowMS int64) {
+	t.Helper()
+	result, err := repo.AdjustLuckyGiftPoolBalance(ctx, domain.PoolAdjustmentCommand{
+		AppCode: "lalu", PoolID: poolID, StrategyVersion: domain.StrategyDynamicV3,
+		AdjustmentID: adjustmentID, Direction: domain.PoolAdjustmentIn, AmountCoins: 1_000_000,
+		Reason: "dual spend integration fixture", OperatorAdminID: 1,
+	}, nowMS)
+	if err != nil {
+		t.Fatalf("fund dynamic spend test pool: %v", err)
+	}
+	if result.Pool.Balance != 1_000_000 {
+		t.Fatalf("funded dynamic spend pool balance=%d want=1000000", result.Pool.Balance)
+	}
+}
+
+func seedClosedDynamicSpendGlobalWindow(t *testing.T, db *sql.DB, rule domain.RuleConfig, nowMS int64) {
+	t.Helper()
+	runtimeConfig, err := luckyRuntimeConfigFromRuleConfig(rule)
+	if err != nil {
+		t.Fatalf("build dynamic spend runtime config: %v", err)
+	}
+	scopeID := luckyRuleWindowScopeID(runtimeConfig)
+	if _, err := db.Exec(`
+		INSERT INTO lucky_rtp_windows (
+		 app_code,scope_type,scope_id,window_index,target_rtp_ppm,control_window_draws,
+		 paid_draws,wager_coins,target_payout_coins,actual_base_payout_coins,carry_ppm,status,
+		 started_at_ms,last_paid_at_ms,closed_at_ms,created_at_ms,updated_at_ms
+		) VALUES ('lalu','pool',?,1,980000,1000,100,1000,980,980,0,'closed',?,?,?,?,?)`,
+		scopeID, nowMS-1_000, nowMS, nowMS, nowMS-1_000, nowMS,
+	); err != nil {
+		t.Fatalf("seed closed dynamic spend global window: %v", err)
+	}
+}
+
+func assertSpendTraceSnapshot(t *testing.T, db *sql.DB, table, keyColumn, key string, wantBefore, wantAfter, wantEarned int64, wantMechanism string) {
+	t.Helper()
+	jsonColumn := "candidate_tiers_json"
+	itemPredicate := ""
+	if table == "external_lucky_gift_draw_items" {
+		jsonColumn = "candidate_snapshot_json"
+		itemPredicate = " AND item_index = 3"
+	}
+	query := fmt.Sprintf(`
+		SELECT
+		 CAST(JSON_EXTRACT(%s, '$.dynamic_v3.pending_spend_jackpot_tokens_before') AS SIGNED),
+		 CAST(JSON_EXTRACT(%s, '$.dynamic_v3.pending_spend_jackpot_tokens_after') AS SIGNED),
+		 CAST(JSON_EXTRACT(%s, '$.dynamic_v3.spend_milestone_tokens_earned') AS SIGNED),
+		 COALESCE(JSON_UNQUOTE(JSON_EXTRACT(%s, '$.dynamic_v3.jackpot_mechanism')), '')
+		FROM %s WHERE app_code='lalu' AND %s=?%s`, jsonColumn, jsonColumn, jsonColumn, jsonColumn, table, keyColumn, itemPredicate)
+	var before, after, earned int64
+	var mechanism string
+	if err := db.QueryRow(query, key).Scan(&before, &after, &earned, &mechanism); err != nil {
+		t.Fatalf("query %s spend trace for %s: %v", table, key, err)
+	}
+	if before != wantBefore || after != wantAfter || earned != wantEarned || mechanism != wantMechanism {
+		t.Fatalf("%s spend trace for %s before/after/earned/mechanism=%d/%d/%d/%q want=%d/%d/%d/%q", table, key, before, after, earned, mechanism, wantBefore, wantAfter, wantEarned, wantMechanism)
+	}
+}
+
 func assertPendingTokenCount(t *testing.T, db *sql.DB, poolID string, userID, want int64) {
 	t.Helper()
 	var got int64
 	if err := db.QueryRow(`
-		SELECT pending_jackpot_tokens FROM lucky_user_states
+		SELECT pending_spend_jackpot_tokens FROM lucky_user_states
 		WHERE app_code='lalu' AND user_id=? AND gift_id=?`, userID, poolID,
 	).Scan(&got); err != nil {
 		t.Fatalf("query pending token: %v", err)
@@ -556,6 +797,9 @@ func TestLuckyDynamicStrategyConfigSeparatesOrdinaryAndJackpotSources(t *testing
 	if runtimeConfig.TargetRTPPPM != 890_000 || strategy.DailyJackpotLimit != 10 {
 		t.Fatalf("v5 runtime target/cap=%d/%d want=890000/10", runtimeConfig.TargetRTPPPM, strategy.DailyJackpotLimit)
 	}
+	if !strategy.JackpotMechanism2Enabled || strategy.MilestoneSpendCoins != rule.JackpotSpendThresholdCoins {
+		t.Fatalf("runtime spend strategy enabled/threshold=%v/%d want=true/%d", strategy.JackpotMechanism2Enabled, strategy.MilestoneSpendCoins, rule.JackpotSpendThresholdCoins)
+	}
 	ordinary := make(map[int64]*domain.StrategyTier)
 	jackpot := make(map[int64]*domain.StrategyTier)
 	for index := range strategy.Tiers {
diff --git a/services/lucky-gift-service/internal/storage/mysql/lucky_gift_dynamic_state.go b/services/lucky-gift-service/internal/storage/mysql/lucky_gift_dynamic_state.go
index dab12c4b..5104ace5 100644
--- a/services/lucky-gift-service/internal/storage/mysql/lucky_gift_dynamic_state.go
+++ b/services/lucky-gift-service/internal/storage/mysql/lucky_gift_dynamic_state.go
@@ -127,6 +127,33 @@ func (r *Repository) getLatestPendingLuckyUserRTPWindow(ctx context.Context, tx
 	return userWindow, global, true, nil
 }
 
+// getLatestClosedLuckyDynamicRTPWindow returns the platform sample used only by the durable spend
+// strategy. It intentionally does not join a user window: a user may have been inactive throughout
+// the latest platform round. Legacy rows without an exact start/close boundary are excluded, because
+// treating an administrative migration close as a settled sample could manufacture jackpot access.
+func (r *Repository) getLatestClosedLuckyDynamicRTPWindow(ctx context.Context, tx *sql.Tx, appCode, globalScopeID string, beforeWindowIndex int64) (luckyRTPWindow, bool, error) {
+	var window luckyRTPWindow
+	err := tx.QueryRowContext(ctx, `
+		SELECT scope_type, scope_id, window_index, target_rtp_ppm, control_window_draws,
+		       paid_draws, wager_coins, target_payout_coins, actual_base_payout_coins,
+		       carry_ppm, status, started_at_ms, last_paid_at_ms, closed_at_ms
+		FROM lucky_rtp_windows
+		WHERE app_code = ? AND scope_type = 'pool' AND scope_id = ? AND status = 'closed'
+		  AND window_index < ? AND wager_coins > 0 AND started_at_ms > 0 AND closed_at_ms > 0
+		ORDER BY window_index DESC
+		LIMIT 1`, appCode, globalScopeID, beforeWindowIndex,
+	).Scan(&window.ScopeType, &window.ScopeID, &window.WindowIndex, &window.TargetRTPPPM, &window.ControlDraws,
+		&window.PaidDraws, &window.WagerCoins, &window.TargetPayoutCoins, &window.ActualPayoutCoins,
+		&window.CarryPPM, &window.Status, &window.StartedAtMS, &window.LastPaidAtMS, &window.ClosedAtMS)
+	if err == sql.ErrNoRows {
+		return luckyRTPWindow{}, false, nil
+	}
+	if err != nil {
+		return luckyRTPWindow{}, false, err
+	}
+	return window, true, nil
+}
+
 func (r *Repository) persistLuckyUserRTPWindow(ctx context.Context, tx *sql.Tx, appCode string, state luckyUserRTPWindow, nowMS int64) error {
 	_, err := tx.ExecContext(ctx, `
 		UPDATE lucky_user_rtp_windows
diff --git a/services/lucky-gift-service/internal/storage/mysql/lucky_gift_dynamic_strategy.go b/services/lucky-gift-service/internal/storage/mysql/lucky_gift_dynamic_strategy.go
index 624bde35..3fbc8174 100644
--- a/services/lucky-gift-service/internal/storage/mysql/lucky_gift_dynamic_strategy.go
+++ b/services/lucky-gift-service/internal/storage/mysql/lucky_gift_dynamic_strategy.go
@@ -131,9 +131,9 @@ func luckyDynamicStrategyConfig(config domain.Config) domain.StrategyConfig {
 		RechargeBoostEndMS:         config.RechargeBoostWindowMS,
 		MissProtectionZeroDraws:    config.LossStreakGuarantee,
 		DailyJackpotLimit:          config.MaxJackpotHitsPerUserDay,
-		MilestoneSpendCoins:        0,
+		MilestoneSpendCoins:        config.JackpotSpendThresholdCoins,
 		JackpotMechanism1Enabled:   true,
-		JackpotMechanism2Enabled:   false,
+		JackpotMechanism2Enabled:   config.JackpotSpendThresholdCoins > 0,
 		GlobalRTPMaxPPM:            config.JackpotGlobalRTPMaxPPM,
 		UserRoundRTPMaxPPM:         config.JackpotUserRoundRTPMaxPPM,
 		User48HourRTPMaxPPM:        config.JackpotUser48hRTPMaxPPM,
diff --git a/services/lucky-gift-service/internal/storage/mysql/lucky_gift_repository.go b/services/lucky-gift-service/internal/storage/mysql/lucky_gift_repository.go
index f5b2a51f..cb646068 100644
--- a/services/lucky-gift-service/internal/storage/mysql/lucky_gift_repository.go
+++ b/services/lucky-gift-service/internal/storage/mysql/lucky_gift_repository.go
@@ -78,9 +78,12 @@ type luckyUserState struct {
 	CumulativeWagerCoins int64
 	EquivalentDraws      int64
 	LossStreak           int64
-	// PendingJackpotTokens 是按 UTC 日消费赚取、但允许跨日等待资金的未消费大奖资格。
-	// 它属于 pool/user 生命周期，不能和“当天消费/当天命中次数”一起在午夜静默清零。
+	// PendingJackpotTokens is the retired token column. Dynamic V3 reads it only so the next paid
+	// transaction can clear it; its value must never become a new spend jackpot qualification.
 	PendingJackpotTokens int64
+	// PendingSpendJackpotTokens belongs to the app/pool/user lifetime. UTC-day spend creates tokens,
+	// but a blocked token survives midnight until one spend jackpot is actually paid.
+	PendingSpendJackpotTokens int64
 }
 
 type luckyCandidate struct {
@@ -2421,11 +2424,13 @@ func luckyCeilDiv(value, divisor int64) int64 {
 func (r *Repository) getLuckyUserState(ctx context.Context, appCode string, userID int64, giftID string) (luckyUserState, error) {
 	var state luckyUserState
 	err := r.db.QueryRowContext(ctx, `
-		SELECT paid_draws, cumulative_wager_coins, equivalent_draws, loss_streak, pending_jackpot_tokens
+		SELECT paid_draws, cumulative_wager_coins, equivalent_draws, loss_streak,
+		       pending_jackpot_tokens, pending_spend_jackpot_tokens
 		FROM lucky_user_states
 		WHERE app_code = ? AND user_id = ? AND gift_id = ?`,
 		appCode, userID, giftID,
-	).Scan(&state.PaidDraws, &state.CumulativeWagerCoins, &state.EquivalentDraws, &state.LossStreak, &state.PendingJackpotTokens)
+	).Scan(&state.PaidDraws, &state.CumulativeWagerCoins, &state.EquivalentDraws, &state.LossStreak,
+		&state.PendingJackpotTokens, &state.PendingSpendJackpotTokens)
 	if errors.Is(err, sql.ErrNoRows) {
 		return luckyUserState{}, nil
 	}
@@ -2438,8 +2443,8 @@ func (r *Repository) getLuckyUserStateForUpdate(ctx context.Context, tx *sql.Tx,
 	if _, err := tx.ExecContext(ctx, `
 		INSERT INTO lucky_user_states (
 			app_code, user_id, gift_id, paid_draws, cumulative_wager_coins, equivalent_draws, loss_streak,
-			pending_jackpot_tokens, created_at_ms, updated_at_ms
-		) VALUES (?, ?, ?, 0, 0, 0, 0, 0, ?, ?)
+			pending_jackpot_tokens, pending_spend_jackpot_tokens, created_at_ms, updated_at_ms
+		) VALUES (?, ?, ?, 0, 0, 0, 0, 0, 0, ?, ?)
 		ON DUPLICATE KEY UPDATE updated_at_ms = updated_at_ms`,
 		appCode, userID, giftID, nowMS, nowMS,
 	); err != nil {
@@ -2447,12 +2452,14 @@ func (r *Repository) getLuckyUserStateForUpdate(ctx context.Context, tx *sql.Tx,
 	}
 	var state luckyUserState
 	err := tx.QueryRowContext(ctx, `
-		SELECT paid_draws, cumulative_wager_coins, equivalent_draws, loss_streak, pending_jackpot_tokens
+		SELECT paid_draws, cumulative_wager_coins, equivalent_draws, loss_streak,
+		       pending_jackpot_tokens, pending_spend_jackpot_tokens
 		FROM lucky_user_states
 		WHERE app_code = ? AND user_id = ? AND gift_id = ?
 		FOR UPDATE`,
 		appCode, userID, giftID,
-	).Scan(&state.PaidDraws, &state.CumulativeWagerCoins, &state.EquivalentDraws, &state.LossStreak, &state.PendingJackpotTokens)
+	).Scan(&state.PaidDraws, &state.CumulativeWagerCoins, &state.EquivalentDraws, &state.LossStreak,
+		&state.PendingJackpotTokens, &state.PendingSpendJackpotTokens)
 	return state, err
 }
 
diff --git a/services/lucky-gift-service/internal/storage/mysql/lucky_gift_rule_config_repository_test.go b/services/lucky-gift-service/internal/storage/mysql/lucky_gift_rule_config_repository_test.go
index c2d9f648..a5244ae9 100644
--- a/services/lucky-gift-service/internal/storage/mysql/lucky_gift_rule_config_repository_test.go
+++ b/services/lucky-gift-service/internal/storage/mysql/lucky_gift_rule_config_repository_test.go
@@ -153,6 +153,14 @@ func TestDynamicStrategyMigrationIsIdempotent(t *testing.T) {
 	// initdb 已包含最终结构，004 连续运行两次仍必须成功，覆盖全新环境与线上重复发布 migration 两条路径。
 	runLuckyGiftSQLScript(t, schema.DB, migrationPath)
 	runLuckyGiftSQLScript(t, schema.DB, migrationPath)
+	// 007 must also work on a pre-dual-strategy schema and remain replay-safe. Dropping only the new
+	// constant-default column reproduces that shape without scanning or rewriting any user row.
+	if _, err := schema.DB.Exec(`ALTER TABLE lucky_user_states DROP COLUMN pending_spend_jackpot_tokens, ALGORITHM=INSTANT`); err != nil {
+		t.Fatalf("prepare pre-007 schema: %v", err)
+	}
+	spendTokenMigrationPath := mysqlschema.InitDBPath(t, caller, "..", "..", "..", "deploy", "mysql", "migrations", "007_dual_jackpot_spend_tokens.sql")
+	runLuckyGiftSQLScript(t, schema.DB, spendTokenMigrationPath)
+	runLuckyGiftSQLScript(t, schema.DB, spendTokenMigrationPath)
 
 	for _, table := range []string{
 		"lucky_risk_counters", "lucky_user_rtp_hour_buckets", "lucky_user_rtp_boundary_events",
@@ -171,6 +179,7 @@ func TestDynamicStrategyMigrationIsIdempotent(t *testing.T) {
 		minLength int64
 	}{
 		{table: "lucky_user_states", name: "pending_jackpot_tokens"},
+		{table: "lucky_user_states", name: "pending_spend_jackpot_tokens"},
 		{table: "lucky_rtp_windows", name: "scope_id", minLength: 128},
 		{table: "lucky_risk_counters", name: "scope_id", minLength: 255},
 		{table: "lucky_gift_command_locks", name: "command_id", minLength: 128},