In balance, shorter exposures, stacking, and track-removal is technically easier.

Long exposures made sense when photgraphic plates were a scarce resource, and replacing them risked disturbing the observation.

Stacking is premised on the idea that individual exposures are cheap, and that noise tends to affect a small number of those exposures, in small regions. The same predictability of satellite tracks you name means that they can be removed through image processing rather than brute force (avoiding sky regions, physical masking, interrupting exposures during overflights, etc.).

And the other phenomena I mention (meteors, cosmic rays) are not predictable, and also degrade deep-space images.