We recently discovered the configuration wsrep_provider_options="pc.recovery=TRUE" and set about to set this configuration, as we'd like to avoid unnecessary boostrapping whenever possible. We were further surprised to discover that this feature had been enabled by default.

But we were frustrated in our attempts to trigger this feature to actually observe it in action.

I understand that the feature is intended to recover from full cluster crashes _such as_ a power outage, but we were unable to come up with a scenario in which we could interrupt the cluster and watch it recover automatically. This included deploying to AWS and setting all three nodes to 'stop' then 'start.' In each instance, we had to manually bootstrap.

Am I misunderstanding the feature, or the conditions under which we can trigger it?