[core] Add the wake-loop scenario to the event log race repro (vercel#4017)

Author: VaguelySerious

.changeset/wake-loop-repro-scenario.md

@@ -0,0 +1,2 @@
+---
+---

.github/scripts/render-event-log-race-repro-results.js

@@ -283,6 +283,7 @@ function renderConfigScale(config) {
     config.blockedBranchAttempts
       ? `blocked-branch ${config.blockedBranchAttempts}`
       : '',
+    config.wakeLoopAttempts ? `wake-loop ${config.wakeLoopAttempts}` : '',
     config.hookSleepAttempts ? `hook-sleep ${config.hookSleepAttempts}` : '',
     // Historical entries from the pre-storm harness, kept so an old sticky
     // comment still renders its own configuration rather than a blank line.
@@ -312,6 +313,9 @@ function renderConfigTiming(config) {
       ? `step ${config.stepDelayMs}±${config.stepDelayJitterMs ?? 0}ms`
       : '',
     config.hookResumeStaggerMs ? `stagger ${config.hookResumeStaggerMs}ms` : '',
+    config.wakeLoopAttempts && config.wakeLoopHeartbeatMs
+      ? `heartbeat ${config.wakeLoopHeartbeatMs}ms`
+      : '',
     config.resumeBurstOffsetMs && config.blockedBranchAttempts
       ? `burst ${config.resumeBurstOffsetMs}+${config.resumeBurstJitterMs ?? 0}ms`
       : '',
@@ -342,6 +346,7 @@ function compactConfig(config = {}) {
     stepStormAttempts: config.stepStormAttempts,
     hookStormAttempts: config.hookStormAttempts,
     blockedBranchAttempts: config.blockedBranchAttempts,
+    wakeLoopAttempts: config.wakeLoopAttempts,
     hookSleepAttempts: config.hookSleepAttempts,
     concurrency: config.concurrency,
     rounds: config.rounds,
@@ -357,6 +362,7 @@ function compactConfig(config = {}) {
     hookResumeStaggerMs: config.hookResumeStaggerMs,
     launchStaggerMs: config.launchStaggerMs,
     resumeBurstOffsetMs: config.resumeBurstOffsetMs,
+    wakeLoopHeartbeatMs: config.wakeLoopHeartbeatMs,
     resumeBurstJitterMs: config.resumeBurstJitterMs,
     blockedBranchWatchdogMs: config.blockedBranchWatchdogMs,
     runTimeoutMs: config.runTimeoutMs,

.github/workflows/event-log-race-repro.yml

@@ -58,6 +58,18 @@ on:
       blocked_branch_watchdog_ms:
         description: 'blocked-branch: hook-race watchdog. Size it to normally lose to the burst; it exists to enter a wait into the race'
         required: false
+      # GitHub caps workflow_dispatch at 25 inputs, so only the wake-loop knobs a
+      # soak is likely to turn are exposed here; the rest are env-only (see
+      # `EVENT_LOG_RACE_REPRO_WAKE_LOOP_*` in the harness).
+      wake_loop_attempts:
+        description: 'Number of wake-loop runs (one sequential loop racing a reusable hook read against a heartbeat sleep; concurrency comes from the driver resumes)'
+        required: false
+      wake_loop_heartbeat_ms:
+        description: 'wake-loop: heartbeat sleep raced against the hook read; part of the wakes are aimed at this deadline'
+        required: false
+      wake_loop_fresh_ratio:
+        description: 'wake-loop: share of wakes that carry work (a stale wake emits no step, a fresh one emits four)'
+        required: false
       poke_interval_ms:
         description: 'step-storm: cadence of out-of-band resumes to the never-read poke hook'
         required: false
@@ -156,6 +168,9 @@ jobs:
           EVENT_LOG_RACE_REPRO_RESUME_BURST_OFFSET_MS: ${{ inputs.resume_burst_offset_ms }}
           EVENT_LOG_RACE_REPRO_RESUME_BURST_JITTER_MS: ${{ inputs.resume_burst_jitter_ms }}
           EVENT_LOG_RACE_REPRO_BLOCKED_BRANCH_WATCHDOG_MS: ${{ inputs.blocked_branch_watchdog_ms }}
+          EVENT_LOG_RACE_REPRO_WAKE_LOOP_ATTEMPTS: ${{ inputs.wake_loop_attempts }}
+          EVENT_LOG_RACE_REPRO_WAKE_LOOP_HEARTBEAT_MS: ${{ inputs.wake_loop_heartbeat_ms }}
+          EVENT_LOG_RACE_REPRO_WAKE_LOOP_FRESH_RATIO: ${{ inputs.wake_loop_fresh_ratio }}
           EVENT_LOG_RACE_REPRO_POKE_INTERVAL_MS: ${{ inputs.poke_interval_ms }}
           EVENT_LOG_RACE_REPRO_POKE_MAX: ${{ inputs.poke_max }}
           EVENT_LOG_RACE_REPRO_POKE_DECAY_FACTOR: ${{ inputs.poke_decay_factor }}
@@ -264,6 +279,9 @@ jobs:
           EVENT_LOG_RACE_REPRO_RESUME_BURST_OFFSET_MS: ${{ inputs.resume_burst_offset_ms }}
           EVENT_LOG_RACE_REPRO_RESUME_BURST_JITTER_MS: ${{ inputs.resume_burst_jitter_ms }}
           EVENT_LOG_RACE_REPRO_BLOCKED_BRANCH_WATCHDOG_MS: ${{ inputs.blocked_branch_watchdog_ms }}
+          EVENT_LOG_RACE_REPRO_WAKE_LOOP_ATTEMPTS: ${{ inputs.wake_loop_attempts }}
+          EVENT_LOG_RACE_REPRO_WAKE_LOOP_HEARTBEAT_MS: ${{ inputs.wake_loop_heartbeat_ms }}
+          EVENT_LOG_RACE_REPRO_WAKE_LOOP_FRESH_RATIO: ${{ inputs.wake_loop_fresh_ratio }}
           EVENT_LOG_RACE_REPRO_POKE_INTERVAL_MS: ${{ inputs.poke_interval_ms }}
           EVENT_LOG_RACE_REPRO_POKE_MAX: ${{ inputs.poke_max }}
           EVENT_LOG_RACE_REPRO_POKE_DECAY_FACTOR: ${{ inputs.poke_decay_factor }}

AGENTS.md

@@ -130,13 +130,18 @@ pnpm run test:e2e
 ### Event log race repro
 
 `packages/core/e2e/event-log-race-repro.test.ts` is a dedicated harness for
-`CORRUPTED_EVENT_LOG`. It drives four scenarios against one deployment:
+`CORRUPTED_EVENT_LOG`. It drives five scenarios against one deployment:
 `step-storm` and `hook-storm` (concurrent replays of a single run racing the
-per-branch watchdog; `hook-storm` is the production shape), `blocked-branch`
+per-branch watchdog; `hook-storm` is a production shape), `blocked-branch`
 (each branch parks on a launch step before its hook race, so a woken replay can
 hold a log that predates a sibling's launch completion and take the ordinal that
 sibling's wait is about to get; it covers the class the wake-order fixes miss),
-plus a `hook-sleep` control that provides the calibration baseline. Any outcome
+`wake-loop` (one sequential loop racing a reusable hook read against a
+heartbeat sleep, no fan-out; the driver supplies the concurrency with bursty
+resumes and resumes aimed at the heartbeat deadline, the shape of a production
+run whose replays of one immutable prefix diverged non-deterministically on an
+unconsumable `wait_created`), plus a `hook-sleep` control that provides the
+calibration baseline. Any outcome
 other than `completed` fails the run, except `infra`, which means the harness
 could not reach the deployment.