There are certainly negative effects of a large transaction, but I thought you want to optimize the performance of a "pg_restore". If you optimize one thing, you will certainly pessimize some other things. In the case at hand, you shouldn't run a heavy data modifying workload in the same database concurrently to the large pg_restore.
pg_restore will be the only thing running.
The mere size of a transaction can be a problem as such in other databases like Oracle, but not in PostgreSQL.
Understood. I'll see if I can run another test before the "real" migration starts.
And... --single-transaction conflicts with --jobs. Oh well.