Normally if one SPO work process is busy or in error the other 2 should take the incomming work.
Can you stop the "waiting" spool requests and start one for test to see if it is going through?
Also you can restart the SPO PID from os level and set the start flag on it in dpmon