before going to accessible-by-default, in addition to tests the API PR must link tested documentation for

Etcd Team to investigate whether a larger etcd database can cause crash-loops in Etcd (via out-of-memory for example) on restarts (similar to the above API server issue).
If it is possible to get into a crash-loop because of the database size, the Etcd team will document how to recover the Etcd cluster from this state.
and if it is "impossible", I'd like to understand where the infinite memory machine comes from.