I keep coming to this issue: the REL exception if you like.
For all the passionate argumentation for and against, there's little raw data on REL in general.
The only proper measurement I've seen of a REL was this one
https://www.audioholics.com/subwoofer-reviews/rel-ht-1510-sub/conclusion
It's for a model in the HT line.
What I saw there clearly looked like the FR of a single driver. Not very flat at any point. More like a parabola / bell curve.
I understood that to be "the REL way" i.e. no DSP applied to the driver, just concentrate on the driver itself, the LP filters and the finish/cabinet.
This appealed to me then.
Been peaking at the multi-sub idea, and the reality that for small rooms, the region <100hz is dominated by discrete modes, and your flat anechoic FR is going to get messed, and is going to be very spatially uneven.
If eventually you are going to do multi-sub, and smooth out via top-level DSP, then I don't see much point to having DSP on each individual sub.
Though having DSP on the sub I guess makes it an easier "building block" for your AVR.
Floyd Toole's book basically says that at low frequencies it's all about getting output "out there".
So, in a way, if what you want are >1 "dumb" individual subs and higher level calibration via mic + DSP,
wouldn't it be preferable to have the sub basically be a driver?
I was eyeing a Kef Kube 10b, and downloaded the instruction manual to see if the DSP had an "off" switch for "just give me the raw driver"... nope.