The problem with running Mawile as a counter to legendaries (and considering how good he is in this case, it's a small problem) is that anyone running Xerneas will be running either HP fire, or a different pokemon to counter Aegislash, and other steels, but mostly Aegislash. HP Fire will only OHKO after Geomancy but Iron Head from MMawile will easily OHKO so as long as you don't swap into it, Mawile still wins.
Second problem is as you mentioned, WoW; Talonflame and Rotom-W (just to name a few) are everywhere from what I've seen in the practice games and easily outspeed Mawile so if they predict your switch, you're screwed, or if you don't switch out of WoW, you're screwed because Garchomp, Kang and MewtwoX are also everywhere and all can eat a burned Play Rough and murder you with EQ.
There is a huge difference between a check and a counter,
please read this article because the distinction is extremely important. You spent a lot of text detailing why Mawile wasn't a counter which I openly admitted, my point was its value comes as a
check. I specifically said Mawile was a check and while yes, all of those things mentioned can beat Mawile....in the most common conditions, Mawile will win. Xerneas and Yveltal require absurdly advantageous situations (and moves with shit accuracy) to win, whereas Mewtwo needs a super specific situation or certain moves, the most common of which was only used on 32% of Mewtwos in the 1695 and 34% in the 1760 stats last month. I think Mawile's a good idea because it requires the least amount of teamslots possible while having the widest beneficial effect; it's not going to win 100% of the time but it will win a lot more than 50%.
This isn't the simulator, HP Fire and Ground Xerneas might as well not exist in-game, the odds are near zero for a competitive one due to the forced 31s making them more difficult to reset for. There is a 50% chance to get a 31 in SpA, SpD or Speed which nullifies all HP Fire spreads (which require 30s in SpA and Speed) or HP Ground (which requires 30 SpD), then factor in the 50% chance to get the desired even or odd Atk, and then the 50% chance to get the nature you want. So at
maximum, there is a 12.5% chance that a specific Hidden Power spread even has a chance to be generated, so 25 out of every 200 resets are even meaningful. I could continue to go into the odds of getting semi-decent IVs in all stats or do proper math with all the other conditional values that I didn't go into for simplicity's sake (i.e. HP Fire needs 30s in BOTH SpA and Speed), but I hope that I don't have to continue to break down statistics to make the point that they are not worth preparing for. If you see one, a regular Xerneas will probably win because the HP Xerneas will likely have some lackluster stat. The potential loss of brain cells and stats aren't worth the situational uses.
Nobody is going to keep Mawile in on Rotom-W and Talonflame, and Mawile is certainly not alone in the not-liking-WoW department. But Mawile's power is such that no, it's not so easy to eat a Play Rough. Talonflame is safe because it resists both of Mawile's STABs, but Rotom-W is going to take a not-insignificant chunk (43.9 - 51.5%), Garchomp can't switch in (83.1 - 97.8%), and neither can Mewtwo X (80.1 - 95%; regular takes 43.6 - 51.3% on the switch-in).
tl;dr: Mawile steamrolls