[jbiskur]

Summary

• New skills/flowcore-pathways/references/resilient-startup.md: production start runtime for virtual pathways. It retries the whole start with stopPump() + stopCluster() teardown, exits after the last attempt, exits on Failed to bootstrap leader runtime after becoming leader (detected in the logger adapter), and runs a leader watchdog. It separates liveness from readiness and adds handler timeouts sized to deliveryTimeoutMs.
• SKILL.md: new rule 11 (a start error must never leave a silent stall). The minimal example now uses a single-attempt startOnce() plus the runtime. Checklist items added.
• cluster-and-pump.md: why the naive start stalls. Fixes the PostgreSQL-outage row: the leader does not step down, because renewLease throws and the error is logged as Lease loop error.
• Troubleshooting (both pathway skills) and a new evaluation scenario.
• Version 0.2.0.

Verification (source)

Every claim was checked against the @flowcore/pathways 2.10.4 source (builder.ts startCluster/startPump/handleLeadershipChange/bootstrapLeaderPump, cluster-manager.ts, postgres-coordinator.ts, pathway-pump.ts) and the @flowcore/data-pump 0.23.3 source (data-pump.ts redelivery and fetch loop). The patterns come from about 40 Flowcore workspace solution and outage fragments on pump stalls.

Test plan

• node scripts/validate-package.mjs passes (13 checks), and the self-test passes 33/33
• deno check of pathways-logger.ts, pathways-runtime.ts, the SKILL.md example and the handler-timeout block against the pathways 2.10.4 source
• Behavior test of the runtime with a fake pathways module:
  • recovers after 2 failed attempts, with teardown between attempts
  • exits 1 after 6 attempts
  • exits 1 on the fatal bootstrap log line
  • the watchdog exits 1 when the leader has no running pump past the grace period
• CI green

🤖 Generated with Claude Code