I don't recall how wide a "critical band" is in the bass region, sorry. My impression is that it's less than 1/3 octave wide.
For those who wonder what I'm talking about, the ear/brain system "averages out" narrow-band peaks and dips across what's called a "critical band", which is approximately 1/3 octave wide over much of the spectrum. So in other words if we have offsetting peaks and dips within 1/3 octave of one another, the ear tends not to hear either one. If we just have a peak, then the ear will hear it, though if it's a narrow peak it looks worse to the eye than it sounds to the ear.
In the bass region, the room-induced peaks and dips are inherently too far apart for the ear/brain system to average them out, and as a result they are audible and objectionable. Higher up we still have numerous peaks and dips from room interaction, but they are so close together (and delayed by so many wavelengths) that they are relatively benign, even though they may look awful on an unsmoothed in-room frequency response curve.
Tieing back in to the multisub concept, each sub will produce a different peak-and-dip pattern so the sum ends up not only averaging out considerably, but the remaining peaks and dips are more numerous and thus closer together - giving the ear/brain system's smoothing mechanism a better chance to work in to our benefit. As a result, the subjective benefit of a multisub system is often greater than one would expect from merely eyeballing the in-room frequency response curves.
It's not necessary to use identical subs like in the system I build; in fact if the subs aren't designed with room gain in mind, we're probably better off with some of the subs extending deeper in the bass than others. The basic principle can be employed without spending megabucks.