There are scheduler-type updates for a lot of things. If you look at a filling or emptying bunch of water, you can see the pattern of updates going over it. Items are checked for for pathability or w/e every 1,2,4,8, etc. up to IIRC 1024 ticks (said IIRC in an interview).
He's done it for water, certainly, because it's such a
massive hit to FPS if he doesn't that he pretty much had to.
It largely amounts to something like "once every 16 ticks", and only tests every fourth tile if memory serves, so it's actually very slow. (Without pressure, dwarves can out-walk a wall of water.) "Still" waters have even less frequent checks, to the point of taking several hundred frames for a tile of water adjacent to a just-mined wall to recognize it now has somewhere new to flow. (Seriously, try mining out a stagnant pool from the side in a test embark, and see how far away your dwarf gets back to the fort before the water starts breaking up from 7s, I've seen as many as 40 steps.)
It doesn't happen for almost anything creature-related, and probably should, as there's little reason to update thirst, hunger, readiness to party and the like between actual turns within which the creature can actually make decisions based upon those properties.
Water, for that matter, could be handled differently. Minecraft, for example, largely only checks when neighboring tiles are changed, rather than periodic polling. (Although Minecraft's system obviously has bugs and serious exploits associated with it, especially since they don't handle flows well or units of water less than a block.) I'd also made a
suggestion a while back on treating water as a whole body, rather than a series of tiles that individually flow, which should also preserve FPS, especially with large bodies like oceans.