I believe it is a good practice that is up to the author, and that is the main part. Up to the author.

I'm not saying "don't do it", I'm saying "don't review it unless the author themselves committed to it". It is not a reason to hold back a PR.

Imagine holding back a feature because you need to re-alias an import to some convention across 10 modules.

It is a slippery slope.

The advantage of a separate PR to re-alias an import is that its very safe, easy to deploy, easy to merge across all working branches etc.

But w/e works for you. "Works" is the best criteria :)