I might of missed it but I don't see anything like that in the ingestion spec.
My Druid version: 0.21.1
I am seeing this odd behavior where sometimes, my index_parallel job would hang forever and that's due to 1 of the sub_task hanging forever.
To overcome that problem, I killed that hanging sub_task and then index_parallel would retry and the problem is solved.
But doing this manually is clearly tedious, if there's a programatic way of solving this (by setting a timeout), that would be great.