XEntropy rdmbair13m5 retry heartbeat

Created a bounded, values-free Codex heartbeat so the one unavailable XEntropy fleet host can converge without disturbing the five verified installations.

Scope

• Controller host: rdmsm4x
• Target host: rdmbair13m5.ts.dataroo.net
• Automation ID: finish-xentropy-fleet-canary
• Schedule: every 30 minutes, bounded to the next eight hours

Exact state and behavior

• Canonical repository: /Users/richh/dev/apps/Xentropy
• Required documentation commit: d5cd6503b636fdb43ca5ad7f15b2265bc1bd9bf1
• Required installed app commit: 6b3261827a064c1389114176b4c1e264992c8636
• The heartbeat first checks reachability and exact canonical state.
• While the host is unreachable it performs no mutation.
• When reachable it inspects the installed checkpoint and, only if needed, invokes the repository's rollback-protected installer for --host rdmbair13m5.
• It sets only the non-secret Tyrell-light UI preference, restarts only XEntropy, and verifies signing, CLI presence, schema-v2 metadata, running state, and the saved template.
• It must stop retrying after a successful canary.

Verification evidence

• Immediate SSH retry at 2026-08-29 06:29 EDT timed out on port 22.
• The existing five-host rollout remains independently verified and unchanged; evidence and rollback locations are in /Users/richh/dev/apps/Xentropy/docs/reviews/2026-08-29-tyrell-light-fleet-rollout.md.

Safety and rollback

• No credential values, provider calls, public listeners, or cross-host authority copies are permitted.
• Any eventual installation uses the installer's per-host rollback directory under ~/Library/Application Support/XEntropy/DeploymentRollback/.
• To stop the retry before its bounded schedule expires, pause or delete the finish-xentropy-fleet-canary automation in Codex.

Outstanding owner action

• None while away. If rdmbair13m5 remains unreachable after the bounded retry window, restore its SSH/Tailscale reachability and rerun the canonical installer for that host only.