Those tactics change now that they know the encounter naturally.
It detects the owl at 120 (most likely) then flies forward using its detect legendary ability to scan for PCs. If it finds any before its turn proper it nukes them. If not, it nukes the (surprised) Owl.
Im not saying its a bad plan mind you. Distraction owls are a valid target.
That said, id be pretty surprised if the familiar lasted this long without death. A party that makes a habit of sending a lone familiar up to scout would have had it killed at the Slaad encounter (and probably a few others), and doesnt have much time to resummon one.
In actual play at a table (with no PC knowledge of the monsters and encounters, and a proactive DM) these encounters would challenge any party.
Id be super intersted to run the experiment again at 17th level. Now THATS a test of 5E. 3E would never hold up at that level.
Good point about the familiar already maybe being dead. The Slaad encounter wouldn't have killed the owl because, it being on the other side of a dimensional barrier, sending through the lone owl is going to be useless. Instead I'd want to send a group of sixteen wolves/whatever, staged in groups of 4, 8, and 4 per round, followed by the PCs on the fourth round. (Their orders are to spread out and defend themselves if attacked. The goal is to trip any ambushes without costing the party any important resources; the wolves may or may not be able to deal with the ambushes all by themselves but they'll sure make it more complicated for any enemies.) I don't remember the encounter description well enough to say if that ideal plan would actually be feasible or if there's some reason to PCs
have to go through alone, but that's the tactical doctrine for jumping into unknown teleportals: send a (disposable) vanguard first if possible.
So the Slaads wouldn't have killed the owl, but something else very well could have. That doesn't change the general doctrine for the atropal fight, it just changes the specifics. If the owl/bat familiar is dead and hasn't been resummoned, push out a recon element of a different kind: summoned animals, an invisible Shadow Monk with a d10 Bardic Inspiration Die and Bless on her, whatever your party can afford. The recon element is going to sweep the room and try to locate any potential threats before bringing in the main body of PCs, so they at least "know" which direction is safe to retreat toward once contact is made (back out of the room? away from the pyramid?
towards the pyramid?). The risk in this case, especially if your recon element is a non-disposable PC, is that there could be traps or hidden threats which cut the recon element off from the rest of the party, e.g. a sliding stone door triggered by a tripwire or pressure plate when you're entering the room. That can make the cautious approach blow up in your face when part of your party has to deal with any threats alone.
Counterplay against that kind of threat is (1) possible; and (2) interesting; and (3) beyond the scope of the current discussion. For now I'll just say the module gets much more interesting if the players are required to be paranoid about both monsters (things you kill via HP attrition) and traps (things you avoid through roleplayed decisions and "utility" spells/abilities) without knowing precisely which they are going to be dealing with next. When I say "your players aren't very paranoid" it should be understood that I mean "in the context of a DM who relies only on monsters." Against a DM who uses both kinds of threats, the exact behavior I'm marvelling at as non-paranoid may in fact be
extremely paranoid, against a different kind of threat.
But, if you're worried about the corridor being cut off, you are still at a minimum going to want to spread out once you're inside the 300' x 300' room--unless you are afraid that spreading out through the bushes/whatnot will trigger hidden tripwires as the clumsier/less observant PCs stumble over them!
[evil DM laughter]