SERVER-126531 introduced a change where we if all log servers are rejected due to stale segment errors (e.g invalid start lsn / truncated lsn errors) from log service we refresh topology from CMS and retry the loop and no longer crash. This addresses the case where we are actually in a recoverable position and the case of a lagged log server. 

However in the case that a standby mongod is lagged too far behind, we need to crash to reload the latest checkpoint, this is for cases where the mongod has lagged beyond the truncation grace period so the log servers can no longer service the reads we want and we need to crash to install the latest checkpoint. 

In order to address this we should bound the retries here instead of continuously retrying, either by retry attempts or bound it by time like the truncation grace period (I think at the moment it is 5 minutes).