Consider the following: 2 files were changed almost simultaneously - A, then B. File A is large enough.
fswatch detects that file A is changed and tells us. We find out that file A was changed and start rsync.
Now, rsync runs for long enough, and during that files B and C are changed too.
After rsync has finished, we set last_synced time to the current time (i.e. to the time when rsync has finished) and find out the next line from fswatch telling us that file B is changed, too.
We check current time and find that just a little moment has passed since we did the last sync (as we check synchronization end time, not synchronization start time). So we don't run rsync.

Now, nothing is changed for a long time. And directories are out of sync: file A is synchronized but file B isn't.
This will only be fixed after some other file is changed, which may happen not very soon.

What we should do IMO: if no new changes happen for rsync_interval seconds but we know about some non-synchronized data then still run rsync.
This will complicate the code logic, and simple for line in ... loop won't work in this scenario. But it will avoid having directories unsynchronized for too long.