Hmmm... interesting concept. Sounds like it could work but depends on, well, several things.
I assume you have to do all your adjustments during preamble time or part thereof, as to not lose anything important. I'm only vaguely familiar with the timing there at the PHY level but that's a pretty short time (the 200us?) and you have to use less of it to leave enough time for the RX station to sync.
You would have to figure out theres a competing station(s) and somehow determine its bearing from you (how, repeated virtual "radar scans" of the virtual antenna?) then calculate and set an antenna array configuration that throws a null on it/them.
You can have an active adaptive antenna array to create nulls in various directions. But, it would be more expensive and complicated and large to do that well. To what level of resolution you need to throw deep enough and narrow enough to be useful nulls, and thereby how complex the active antenna array, I dont know. Nulls arent infinitely deep, maybe -15-30db or so? Tighter aiming and deeper nulls need more complex array. There could be stations at too close of a bearing for a null to cancel practically, you're just out of luck in those cases.
Sounds like a good idea that could work, but I wonder abt practicality, and pricing. I've gone out on a limb pretty far here w my top of head guessing, but I'd think you probably wont see it in >$100 routers, soon.