A batch of reliability fixes: workflows survive a restart, failures surface instead of disappearing, and ready-ticket digests actually send.
- Workflows kept running after a restart. A workflow could be scheduled but no longer runnable, and the failure was silent — it now recovers on demand.
- Failures are reported. A workflow that fell over in a way nothing was watching for used to leave no trace at all.
- Ready-ticket digest emails send again. The query behind them had been failing on every run since 29 July, so no digest was ever delivered.
- Long inbound emails no longer break the reply workflow. A very long thread could exceed the size limit on the way in and fail the whole run.