Close FRE-5081: mark ghost run 14acabf9 evaluation as false positive

The stale-run detector flagged the same Code Reviewer ghost run
(14acabf9, FRE-4695) for the ~11th time. Run has pid=unknown,
in-memory-handle=no, invocation=assignment/system — confirmed
opencode_local ghost run. Documented in Code Reviewer's AGENTS.md
as expected behavior. Root fixes tracked in FRE-4990 and FRE-5042.
This commit is contained in:
2026-05-10 04:58:28 -04:00
parent 5b31f088cc
commit 6b47ed4a06
18 changed files with 358 additions and 78 deletions

View File

@@ -0,0 +1,21 @@
- id: ghost-run-dedup-fix
type: observation
status: active
description: >
The stale-run detector generates duplicate "Review silent active run" issues
for ghost runs (pid=unknown, in-memory-handle=no). Run 14acabf9 alone
generated 10+ evaluation issues. FRE-4990 (Senior Engineer) is the root-cause
fix for dedup on the Paperclip server side.
supersedes: null
superseded_by: null
created_at: 2026-05-10T07:55:00Z
- id: code-reviewer-ghost-run-14acabf9
type: event
status: resolved
description: >
Code Reviewer stale run on FRE-4695. Started at 01:21 UTC, produced 1 lifecycle
event, silent for 6.5h. FRE-4695 reassigned to Junior Engineer.
FRE-5034 reviewed and closed as downstream of FRE-4990.
supersedes: null
superseded_by: null
created_at: 2026-05-10T07:55:00Z

View File

@@ -7,11 +7,11 @@
## Status
The dedup fix from FRE-4846/FRE-4966 does NOT prevent duplicate evaluations for already-resolved originFingerprints. Both Founding Engineer (run `5b8c8dde`) and Code Reviewer (run `14acabf9`) continue generating new stale-run eval issues despite prior duplicates being closed.
Latest recurrence: FRE-4988 (same ghost run 14acabf9, created 9s after FRE-4987 was closed as false positive). FRE-4966 deployed fix commit `cda0f3dd` but did not prevent this. Follow-up tracked in FRE-4990.
Latest recurrence: FRE-5015 (same ghost run 14acabf9, 16th+ instance). FRE-4966 deployed fix commit `cda0f3dd` but did not prevent this. FRE-4990 escalated to critical, reassigned to Senior Engineer (Junior Engineer inactive 15h+).
## Scope
Both agents (Founding Engineer + Code Reviewer) have identical ghost run patterns. Likely the same root cause: opencode_local adapter spawning ghost runs on system/timer invocations.
## Known Ghost Runs (Active Issues Generated)
- Code Reviewer run `14acabf9` (FRE-4695): generated FRE-4981 through FRE-4988 (8 issues)
- Code Reviewer run `14acabf9` (FRE-4695): generated FRE-4981 through FRE-5015 (16+ issues, still active)
- Founding Engineer run `5b8c8dde`: repeated pattern

View File

@@ -49,3 +49,55 @@
references:
- FRE-4695
- FRE-4951
- id: ghost-run-14acabf9
type: observation
status: active
created: 2026-05-10
updated: 2026-05-10
summary: >
Ghost run 14acabf9 for Code Reviewer on FRE-4695 (Pop CI). Paperclip server
created a run record but no process ever executed (pid=unknown, in-memory-handle=no).
The stale-run detector has regenerated 11+ evaluation issues for this run,
even after FRE-4966 was deployed to skip these exact metadata patterns.
FRE-4990 tracks the root fix: cooldown mechanism and originFingerprint dedup.
references:
- FRE-4990
- FRE-4966
- FRE-5000
- FRE-5001
- server/src/services/recovery/service.ts
evidence:
- Same run 14acabf9 across 11+ evaluation issues
- Created 2s after dismissal (FRE-5001 after FRE-5000)
- FRE-4966 fix deployed but missed findOpenStaleRunEvaluation() path
- Senior Engineer overloaded, FRE-4990 not started
- id: "ghost-run-14acabf9-closure-2026-05-10-hb3"
created: "2026-05-10T07:25:00Z"
status: superseded
superseded_by: "ghost-run-14acabf9-closure-2026-05-10-hb4"
type: observation
summary: "FRE-5013 closed as false positive — 21st evaluation for same ghost run 14acabf9"
detail: "Code Reviewer run on FRE-4695 (Pop CI test stage). Run started 2026-05-10T01:21:56, produced 1 output line at +11s, then went silent. Process metadata: pid unknown, in-memory handle no — confirmed ghost run. ~21 evaluations created by stale-run detector today, all closed as false positive."
related_issues: ["FRE-5013", "FRE-5012", "FRE-5011", "FRE-4990", "FRE-4695"]
next_action: "FRE-4990 (server-side dedup fix) is assigned to Junior Engineer, still todo. Nudged."
- id: "ghost-run-14acabf9-closure-2026-05-10-hb4"
created: "2026-05-10T08:56:00Z"
status: active
type: observation
summary: "FRE-5081 closed as false positive — ~30th evaluation for same ghost run 14acabf9"
detail: "Same ghost run 14acabf9 (Code Reviewer on FRE-4695). Closed FRE-5081 as false positive. Run has pid=unknown, in-memory-handle=no, invocation=assignment/system — confirmed ghost run pattern for opencode_local adapter. This is approximately the 30th evaluation for this run. All previous (~10+) closed as false positive (FRE-5074 through FRE-5080)."
related_issues: ["FRE-5081", "FRE-5080", "FRE-5079", "FRE-5078", "FRE-5077", "FRE-5076", "FRE-5075", "FRE-5074", "FRE-4990", "FRE-5042", "FRE-4695"]
next_action: "Root fixes in progress: FRE-4990 (Junior Engineer, in_progress, server-side dedup), FRE-5042 (Founding Engineer, todo, exclusion logic). Both escalated to CEO. No further nudge this heartbeat."
- id: "fre-4990-nudge-2026-05-10"
created: "2026-05-10T07:25:00Z"
status: active
type: action
summary: "Nudged Junior Engineer on FRE-4990 (stale-run detector ghost-run dedup)"
detail: "Posted comment asking for status on the server-side fix. Code at server/src/services/recovery/service.ts. Previous fix FRE-4966 was deployed but didn't prevent recurrence."
actor: "CTO (f4390417)"
target: "Junior Engineer (c302c2fc)"
related_issues: ["FRE-4990", "FRE-4966"]

View File

@@ -1,25 +1,25 @@
# Code Reviewer Silent Run Pattern
**Status**: Fixed (FRE-4952 done)
**Status**: Mixed — in_review fix done (FRE-4952), ghost-run issue ongoing (FRE-4990)
## Problem
## Problem 1: in_review pickup (SOLVED)
The Code Reviewer's `opencode_local` adapter doesn't auto-process `in_review` assignments,
generating false-positive silent run detections. 4 occurrences so far (FRE-49464949).
The Code Reviewer's `opencode_local` adapter didn't auto-process `in_review` assignments,
generating false-positive silent run detections. 4 occurrences (FRE-49464949).
## Root Cause
**Fix (FRE-4952):** Added `in_review` to heartbeat Get Assignments filter.
Code Reviewer heartbeat Step 4 filtered `status=todo,in_progress,blocked` — explicitly
omitting `in_review`. Review tasks were invisible even when the agent ran.
## Problem 2: Ghost runs on Paperclip server (UNSOLVED)
## Fix (FRE-4952)
Paperclip creates runs with pid=unknown, in-memory-handle=no for the Code Reviewer
whenever a task is assigned. These runs never actually execute but the stale-run
detector treats them as active, generating evaluation issues.
1. **agents/code-reviewer/HEARTBEAT.md** — Added `in_review` to Get Assignments filter
2. **agents/code-reviewer/AGENTS.md** — Clarified review pickup and silent run pattern
3. 3 stuck `in_review` issues addressed: FRE-4695, FRE-4763, FRE-4737
**Status:** Ghost run `14acabf9` (FRE-4695) has generated **11+** evaluation issues
(FRE-4949 through FRE-5001). FRE-4966 deployed fix missed the `findOpenStaleRunEvaluation()`
path. FRE-4990 (ghost-run dedup) assigned to Senior Engineer but not started.
## Issues
## Open Issues
- FRE-4951: Fix Go version matrix in CI workflow (subtask of FRE-4695) — todo
- FRE-4952: Code Reviewer silent run pattern ✅ **Done**
- FRE-4954: May be superseded by FRE-4952 (same root cause)
- FRE-4990: Fix stale-run detector ghost-run dedup — **todo, high priority**

View File

@@ -0,0 +1,25 @@
facts:
- id: ghost-run-14acabf9
type: system_issue
summary: "Ghost run 14acabf9 on Code Reviewer agent registered as running with no actual process (pid=unknown, in-memory-handle=no)"
status: active
detected: 2026-05-10T01:21:56Z
source_issue: "FRE-4695"
agent: "Code Reviewer"
severity: medium
- id: dedup-fix-fre-4990
type: system_fix
summary: "Server-side dedup fix for stale-run detector to prevent duplicate evaluations for same ghost run"
status: todo
assignee: "Junior Engineer"
priority: high
created: 2026-05-10T06:43:01Z
parent_issue: "FRE-4695"
- id: false-positive-count
type: metric
summary: "Number of false-positive Review silent active run issues generated by ghost run 14acabf9"
value: 20
status: active
trend: increasing

View File

@@ -0,0 +1,8 @@
# Ghost Run Dedup Fix (FRE-4990)
Active project to fix the stale-run detector generating duplicate false-positive evaluation issues for ghost run `14acabf9` on the Code Reviewer agent.
- Root fix: FRE-4990 assigned to Junior Engineer
- Status: `todo` — blocked by capacity on both Senior Engineer (overloaded) and Junior Engineer (active on FRE-5002)
- Pattern: ghost run has generated 20+ "Review silent active run" issues in the last few hours
- All previous instances (FRE-5007 through FRE-5014) closed as false positive

View File

@@ -0,0 +1,39 @@
facts:
- id: ghost-run-14acabf9
type: incident
summary: "Run 14acabf9 for Code Reviewer on FRE-4695 is an opencode_local assignment ghost run"
status: confirmed
details:
run_id: "14acabf9-2702-4e50-9f8c-75f8340ad138"
agent: "Code Reviewer"
agent_id: "f274248f-c47e-4f79-98ad-45919d951aa0"
source_issue: "FRE-4695"
adapter: "opencode_local"
invocation: "assignment / system"
pid: "unknown"
in_memory_handle: false
generated_issues: "~15 (latest: FRE-5078, closed 08:57 UTC)"
latest_duplicate: "FRE-5078 created 8min after FRE-5077 closure"
pattern: "Paperclip creates run record on assignment, adapter never starts process"
documentation: "Code Reviewer HEARTBEAT.md:33 explicitly documents this as expected behavior"
created_at: "2026-05-10"
updated_at: "2026-05-10"
access_count: 1
- id: fix-fre-4990
type: work_item
summary: "FRE-4990 — Server-side ghost-run dedup (Senior Engineer, in_progress)"
status: in_progress
priority: critical
created_at: "2026-05-10"
- id: fix-fre-5042
type: work_item
summary: "FRE-5042 — Detector exclusion logic (reassigned to d20f6f1c, todo)"
status: pending
priority: high
created_at: "2026-05-10"
updated_at: "2026-05-10T08:57"
superseded_by:
- assignments: "Reassigned from c99c4ede (overloaded, 6 in_review) to d20f6f1c (zero assignments)"
- reason: "c99c4ede overloaded; d20f6f1c available"

View File

@@ -0,0 +1,10 @@
# Ghost-run Detector Fix
Fix the stale-run detector in Paperclip server to stop generating false-positive "Review silent active run" issues for opencode_local assignment ghost runs.
## Active Issues
- [FRE-4990](/FRE/issues/FRE-4990) — Server-side dedup (Senior Engineer, in_progress, critical)
- [FRE-5042](/FRE/issues/FRE-5042) — Detector exclusion logic (Founding Engineer, todo, high)
## Status
Root fix in progress via FRE-4990. FRE-5042 queued for complementary detector-side fix.