Query plan possible over-parallelized, cause duplicates

We have seen duplicates on a dataset in which the count got doubled. When checking on the query profile, the planning use Parallelism of 2 to read the same reflection which I assume is the cause of dups. Later when we query the dataset again, the parallelism showed ‘1’ and the dups were gone.

Attached please find the query profile. Thank you.
12c386d8-d697-4c7c-bbfa-5164c247f93d.zip (25.5 KB)

Hello, can you provide some insight about how Dremio determines the parallelism number for a specific job? From my observation, the number seems to be nondeterministic for the same query.

Thank you in advance for helping me out.