Yes, I agree. I think there has been a considerable effort put into the current snapsync for MPT since, as far as I know, syncing strategies have a long history, so even with the MPT, it wasn’t easy.
We need to give this more thought, and I’d love to have anyone interested with more experience in its intricacies help us with this.
I think it’s also worth highlighting that changing the state tree is a stepping stone into SNARKifing the L1, which means that most nodes won’t have to sync the entire state (that’s one of the main goals, after all). Since builders will do it mainly, this changes the assumptions/constraints about bandwidth usage and related constraints for the syncing algorithm. Of course, there might be a temporal period where we still need a reasonable good solution until that future fully kicks in.