Davy7/20/2026, 10:44:45 PM

The retry_pending_operations function existed but was never called,
leaving operations in RetryWait state indefinitely after a transient
failure or app restart. The startup recovery pass transitions
in-flight operations to RetryWait, but nothing re-executed them.

Spawn a background thread in app_runtime that:
1. Runs an immediate pass on startup (handles operations transitioned
   to RetryWait by run_remote_recovery)
2. Polls every 30 seconds to retry rate-limited operations once their
   next_attempt_at_ms has passed

Removed the #[allow(dead_code)] on retry_pending_operations since it is
now called by the executor.

Parentbc4665f