Seen that happen several times with extremely large searches. Probably due to serious inefficiencies in the execution/serialization of results. Not much you can do about it other than chunk the results yourself; add your own filters in

getInputData

to reduce the result set and set up multiple deployments to process different chunks