I’ve run a 2.2 setup for years now. Subs are funny beasts as in complicated. They interact with the room far more than we can imagine. But at the same time is massive advantage as the location for <80hz subs and speakers is going to be quite different physically to work with the room. Full range loudspeakers cannot be optimized because of this. Usually the sub response will suffer in favor of stage, imaging and presence.
Pointing or facing the subs in a particular direction should not make an audible difference. If it does I reckon the crossover frequency is too high.
Multiple subs have the advantage of having to do less work each but more importantly smoothing the standing waves that develop in anything but open space or very very large space.
Getting the individual phase correct for each sub at the single listening spot is foundational. Hopefully your subs have a phase dial and not just a switch to ‘flip’ phase. One can tame constructive peaks but no amount of power can fill in the rooms natural nulls. Moving the listening location within the space is the only proper solution to nulls. Aka you don’t want to be in a null. Same with doing the sub crawl, adjust the listening location by a couple inches either direction and listen.
One of the biggest bangs for the buck for my system was working with Thierry at Home Audio Fidelity to develop convolution files using mic sweeps in my space. Most of the work is done in the sub frequencies, 80hz and below but can also extend to and improve in room response north of there.
Another thing I’ve found made a tremendous improvement is running the subs via high-level signal from my main power amp driving stand mount speakers. This ensure at least a time coherent signal reaches the sub. Using the low-level input(s) on the subs introduces other problems that one can avoid by driving direct from main power amp. A transformer can be used to convert the high level output to low level if the subs don’t accept speaker level directly. Jensen makes every flavor to fit nearly every combination of input and output.
I would recommend running speaker level into subs and getting phase correct before venturing down the convolution path as it will make the job significantly easier to correct in the time domain…
Hopefully this gives you some food for thought on your journey to the holy grail of flat, in time/phase, sub response.