Silent exception swallowing in mcp/server.py:485-499 (spool drain + retention scheduler)

[emp3thy]

Problem

mcp/server.py:485-499 wraps spool.drain() and the retention scheduler call in broad excepts that swallow exceptions silently with no logging. If either silently fails for weeks, no one will know — the retention scheduler in particular is invisible (it runs in the background of the next retrieve call).

Suggested fix

Route these exceptions into the hook_errors table the way hooks/_error_log.py already does, or at minimum log via logging.exception on a named logger. The point is that a failure should produce a discoverable diagnostic, not vanish.

Acceptance criteria

• Exceptions in the spool drain and retention scheduler paths are recorded somewhere queryable (DB row or log file)
• A test covers the failure-recorded behaviour for at least one of the two paths