The other possible explanation is that it was perfectly possible to run UNIXes
of that era (V4 on) on machines without enough main memory to hold the kernel
and a 'full-sized' process simultaneously. (Our original machine, an -11/40,
started out without a lot of memory; I don't recall exactly how much, though.
It had, I'm pretty sure, 3 banks of core; I was thinking it was 3 MM11-L core
units, which would be 3x16KB, or only 48KB, but my memory must be wrong;
that's not really enough.)
It's interesting - we did the same thing at CMU - almost all of the 11/40's, 40e's and /34's we had in the 1970s running UNIX. We would order them with very minimalist - i.e. 48K and run UNIX swapping to RK05's (slowly) until we had the money to buy more memory and rotating storage ( which was almost always aftermarket - not from DEC) and then those systems would be maxed to 256K. Mellon actually got an original RK07 cheap, so its disks were all Digital (tape was aftermarket), but most of the Unix boxes had dual RK05's and then some storage that we could get cheap. A couple of years later, after I graduated and moved on, I think many groups put Enable's in a couple of them so they could break the 256K barrier - as I sent that code back to Ted and Mike after I wrote it at Tektronix.