Fix: PE: Re-initiate _active_ recurring monitors that previously failed but have timed out

...but do not mark the resource as failed.
This is particularly relevant for those with on-fail=block which stick
around and are not cleaned up by a subsequent stop/start