Memory alert fires but autoscaler does not restart container on Scalingo

Determine whether Scalingo’s autoscaler can be configured to automatically restart containers that repeatedly trigger the platform’s memory‑usage alert (exceeding 80 % of allocated RAM for more than five minutes).

The default autoscaler does not initiate a restart based solely on this alert, leaving the decision to manual intervention or custom health‑check scripts. It is unclear whether a custom health‑check defined in scalingo.json or another platform setting can change this behavior, and what constraints (such as stateful data preservation or billing impact) would apply if automatic restarts were enabled.

Can the autoscaler be instructed to restart a container after a specified number of consecutive memory alerts? Does defining a custom health‑check in scalingo.json override the default no‑restart behavior? Are there any risks or limitations to enabling automatic restarts for stateful applications on Scalingo?