Ignoring based on file metadata becomes weird and complex and I’m strongly biased against implementing it.
Consider a file that is just at the boundary between being ignored and not, and is updated now and then on two sides. Sometimes it’s ignored on one device, sometimes on the other, sometimes we ignore updates from the other side and sometimes we don’t. What does it mean if our ignore cutoff is 10 MB, the file is currently 9 MB, and the other side sends an update at 11 MB – do we ignore that update but continue syncing the file with other devices? Do we sync the update and then ignore the file? What about if it shrinks again? And so on. I wouldn’t look forward to trying to explain this behavior to someone.
There’s a similar argument here: