ARC uses a fine grid although it deduced that it shouldn't

[alongd]

Describe the bug
Got the following in the log file:

  Ending job opt_a62406 for HNNO (run time: 0:35:17)
Error: Optimization job for HNNO with a fine grid terminated successfully on the server, but crashed during calculation. NOT running with fine grid again.
Running local queue job opt_a62438 using gaussian for HNNO (fine opt)

How to reproduce
Run [O-][N+]=N at B2PLYPD3/aug-cc-pVTZ

Expected behavior
Continue without a fine grid if necessary, or trsh in another way (?)
Seems like trsh keywords are not transmitted well into the Scheduler