Looking at the events log. two different timers execute practically simultaneously. Both announce themselves with logWarn “fired”.
Then it looks like one gets to the next logWarn “New value”, but the other timer blows up with a null error before it gets to that point.
Of note, the map-array doesn’t get used in the timer code before “New value”.
I don’t think there is firm doctrine on this, but I believe concurrent access of Group members while updates are going on can give null results, so I’d point the finger at the .members action. Rich @rlkoshak may have a view here.
If it were me, I would put a queuing lock on the timer execution block.
But it might be worth trying the other direct registry method to retrieve members.
Be interested to know if that helps - my own code has .members.findFirst etc. with locks, be useful to know if direct registry read was more robust.