See #79215.

We need a good reproduction for a situation in which a follower is overloaded via incoming raft messages. We need to set this up in a way that cleanly separates the effect of the overload on indirect foreground traffic (i.e. writes reaching the follower as part of quorum, but with the lease elsewhere) and direct overload (i.e. reads/writes for which the overloaded node holds the lease).

This can then be used to prototype improvements to the current behavior.

Jira issue: CRDB-16089