Steady memory growth / suspected leak in self-hosted gateway v2.11.1 after upgrade

After upgrading the self-hosted gateway to `v2.11.1`, pod working-set memory climbs steadily and linearly under steady traffic with no plateau. The configuration is unchanged from the previous version, which held a stable steady state. CPU stays flat throughout, so this is retention growth rather than load-driven, pointing to a regression in 2.11.1.

```
resources:
limits:
cpu: 4
memory: 20Gi
requests:
cpu: 251m
memory: 512Mi
```

Is there a known memory regression in 2.11.x, or a change in 2.11.0/2.11.1 that could account for this growth?

Start by reproducing the steady-traffic workload on gateway v2.11.1 and comparing its pod working-set memory with the previous version. Review changes in 2.11.0 and 2.11.1 for retention regressions; done means the cause is identified and the gateway reaches a stable memory plateau under unchanged traffic and configuration.

Written by the indexing model from the issue text.