I think this was mentioned in another feature request.

Snap to road is something that I want to add, but I want it to be automatic, and something that’s taken into account only partially, at the recording engine level.

You wouldn’t want snap to road for something like walking, because roads aren’t a great indicator of walking routes. But for something like train trips, snap to “road” would be ideal. So I think it should be the recording engine and classifier engine’s jobs to decide how much snapping to use, and when.