Ensure ACL replication isn't too stale

We could check the last contact time for the ACL replication RPCs to ensure they aren't too stale. If we detect that they are outside of some threshold (which probably should be configurable) then we should reissue the query as non-stale or to a different server.

This would all be to ensure we have reasonably up to date ACL data.

Start by tracing the ACL replication RPCs and how their last contact time is tracked. Determine how a configurable staleness threshold and non-stale or alternate-server queries should fit the existing replication flow; the work is done when ACL data is kept reasonably up to date and the behavior is covered by relevant tests.

Written by the indexing model from the issue text.