Dremio goes stale when a dataset is refreshed through API.
Not able to cancel that particular refresh request finally ending to restart the cluster.
Please attach the profile for the reflection refresh job and we can take a look.
I cannot get the failed job profile, but I can attach job profiles of same job which works super fine (completes in less than a second) in Dremio 3.1.3 Dremio_3.1.3.zip (3.2 KB)
and in 3.2.2 Dremio_3.2.2.zip (3.3 KB) it takes a minute and a half to complete.