This was implemented network-wide. The rationale is explained in this meta.SE question:

http://meta.stackexchange.com/q/149722/203101

Briefly:

> If it's really too minor, reviewers should demonstrate that by providing a not-minor edit. If the reviewer opts to build upon the edit instead of starting over from the current revision, then it isn't too minor!

So, improve the edit if it can be improved and reject if it doesn't help at all.