v0.61.9

CRITICAL fix. Agent updated to 0.61.6; update it (hosted fleets self-update).

Fixes (#127)

• A failed update no longer leaves the site offline. WordPress writes a .maintenance flag during an upgrade and deletes it at the end, but several failure/early-return paths (and a failed rollback) skipped the delete, leaving the whole site serving HTTP 503 to every visitor until the file was removed by hand. The agent now guarantees the maintenance flag is cleared on every terminal outcome of an update and a rollback (success, failure, exception, timeout) via a try/finally plus a shutdown backstop, and proactively heals a stale flag left by a prior incident on the next managed update.
• A normal migration 503 no longer triggers a needless rollback. The post-update health probe was a single immediate request, so a plugin that briefly returns 503 while running a database migration on activation (e.g. a major-version bump) could make a successful update look unhealthy and get rolled back. The probe now retries with backoff (~21s window) and only rolls back if the site is still unhealthy after all attempts.

Upgrade

• Pull ghcr.io/mosamlife/wpmgr-{api,web,media-encoder}:v0.61.7 and update the agent to 0.61.6.