I think the best metric is to sort by number of comments on a particular changeset. More comments, more wide spread discussion, more people involved => more impact.

WireGuard for example is only a relatively small changeset, but has an huge impact for many uses, you can find many comments, posts, write ups, about it all over the internet. Many knew that this feature was going to be huge, well before merge into linux.

On the other hand, if almost no one is discussing a particular change (e.g. the removal of 20k lines of driver code), then the impact will either be very low and unimportant or it will unexpectedly become important, but there was no way you could have known it beforehand.