Hey, if anyone can offer a photo of the keyboard innerds (thanks in advance to NTSC, whether or not it ends up being feasible) to reverse-engineer, I'd be one heck of a leg up from where I am now. Even if I figure out the interface protocol, it's still trial-and-error to figure out how all the "specially labelled" keys map to whatever addresses or matrix-locations the CPU is looking for, so any info I can get is helpful.
Regarding the notion of it just being a very simple wire-matrix breakout (ala Alex Eisenhut's comment, and NTSC as well) you know I hadn't thought of that, but now that you mention it, yeah - it does seem like the very sort of thing you WOULD see in a late 1970s or early 1980s machine. Somehow I guess I'm just accustomed to expecting "elegant" solutions whether they were used or not. But, yeah, it sure would explain the need for a large pin count on the connector, and it makes WAY more sense than something silly (like setting up a custom system with parallel data lines and a large word size just to talk to a keyboard). And yes, the scan rate would probably be in the kHz (or maybe less), so it would be REAL easy to see the "polling" happen. I gotta try that out...
And maybe my memory is failing me, but I vaguely remember peeking inside an old "numeric keypad" for an Atari once, and being shocked to find just a switch matrix (with NO ICs) mapped out direct to the connector. This is a bit fuzzy, but if you say that some Commodores did the same kind of thing, I'd totally believe it. Hey, thanks for the clear-headed perspective!