configuration-guide.md has no section explaining how the consent-expiry sweep or the webhook delivery worker behave when the accelerator is deployed across multiple IS nodes. Both jobs run independently per node and rely on a DB-level claim (conditional UPDATE/DELETE) for correctness — reviewed and confirmed safe in #119 and #127 — but that design isn't written down anywhere for operators.
Add a short section covering:
- Each node runs its own timer; there's no leader election or single "active" node.
- Per-node config values (batch size, worker-pool size, poll/interval seconds) are per instance — cluster-wide throughput scales as
N × value, not a shared cluster-wide setting.
- No manual clustering setup is required (unlike the old Quartz-based job).
configuration-guide.mdhas no section explaining how the consent-expiry sweep or the webhook delivery worker behave when the accelerator is deployed across multiple IS nodes. Both jobs run independently per node and rely on a DB-level claim (conditionalUPDATE/DELETE) for correctness — reviewed and confirmed safe in #119 and #127 — but that design isn't written down anywhere for operators.Add a short section covering:
N × value, not a shared cluster-wide setting.