Catch corrupted output files from WriteBuffer failure jobs.
In a previous production job B2OC/AnaProd-B02DKPi/v0r0p2518507/2017_13364401_MagDown
one of the output files was corrupted due to (probably) a lack of disk space/failing disk on the worker node. This resulted in 1000s of lines of Error in <TBranch::TBranch::WriteBasketImpl>: basket's WriteBuffer failed.
in the DaVinci job for one of the constituent files.
The only remedy for this is to just resubmit the job in a later production so perhaps it should be caught during the job processing to rerun the job then rather than after a resubmit. This seems to be quite a rare issue, for me it has happened once out of a few thousand files, but should still be prevented if possible.
Attached file is the DaVinci log for the corrupted file. DaVinci_00140874_00000004_1.log