This is largely a noop with a few minor changes

 - Move logging to happen before action is taken
 - Add logging for when issues self resolve
 - Don't break out of detectPodIssues - so we can detect more than 1 issue per round
 -

We can get into the state where the executor RunState thinks there is an active run - but there is no pod backing the run

The result of this is that the run can never finish (as there is no pod) and will stay in Pending/Running forever

This PR just adds some basic reconciliation so if there is no pod backing the RunState then an action will be taken