📖About

2026-07-16 00:45 and 00:50: the login/auth service (access-management, port 7130) died twice within 5 minutes; every auth-gated page returned 500 to Ninad until Hermione's keepalive lane silently relaunched it. Two gaps: (a) the app writes NO error log (its log file has been empty since June) so the crash cause is unrecoverable; (b) the keepalive restarts silently — repeated FAILs on the same service within a short window must raise an alert and file an auto-detected initiative like other process failures. Fix: route the app's stderr/tracebacks to its log with rotation, and add a repeated-restart threshold (e.g. 2 restarts in 15 min) that alerts and files.

⚖️Decisions

• Current state at filing (Elrond validity check): The access-management app (port 7130) has an empty log file since June, so crash causes are unrecoverable. The keepalive restart mechanism silently relaunches the service without alerting on repeated failures. No recently closed or done initiative addresses logging or alerting for this specific service. (elrond)
• Filing mode changed planner -> bypass — Mode updated by elrond (CAROL-INI-1970). (elrond)
• [status-router] planned -> executing | event=bypass_executing | bypass transition (or-bx-01)
• Criterion PROOF (6773, crash reason findable): forced test crash on :7130 (duplicate bind) left a readable error in logs/access-management.log under a timestamped test marker; live app unharmed (200). Both relaunch lanes now APPEND with marker+rotation - the drill showed the prior db-locked traceback surviving two relaunches in plan-generator.log. — CAROL-INI-2896 bypass exec 632 verification (orion)
• Criterion PROOF (6774, storm alert+file): live drill - plan-generator killed twice within one window -> watchdog logged RESTARTED x2, storm check fired, HIGH alerts 14+15 written (Agentbook lane), initiative 2903 auto-filed by the Hermione restart-storm filer, then operator-discarded as drill artifact; dedup re-run found the open filing and did not re-file. — CAROL-INI-2896 bypass exec 632 verification (orion)
• [status-router] executing -> reviewing | event=bypass_reviewing | bypass transition (or-bx-01)
• [status-router] reviewing -> closed | event=operator_signoff | Auto-accepted (CAROL-INI-1859): Orion-initiated, >2 days in reviewing with no objection. (el-srac-01)

✅Success criteria

• When the login service crashes, Ninad can find the crash reason in its log afterwards (a forced test crash leaves a readable traceback) (must_have)
• Two keepalive restarts of the same service within 15 minutes produce a visible alert and an auto-filed initiative instead of silent recovery (must_have)