you said,
"... for multiple presses with low on selected out you get multikey, with high on selected out, you get no key. The 'no key' most likely will be preceeded as well as followed by a key read."

I don't follow ... the value read from the row inputs is zero when no key is pressed. It's nonzero when there's a key closure, even a multiple key closure. Presumably, the high-resistance pulldown won't compromise its low-resistance high generated by the column select. I can see why one would prefer a low with the 805x output port structure, but, this thing was a dedicated (DEC) keboard encoder IC of yesteryear. It drives the "columns" high, one at a time, and looks for the high to appear at its "row" inputs. I fail to see why it makes any difference whether the presumably, but not necessarily, unique input is a high against a background of lows, or a low against a background of highs.

Was this choice by the IC manufacturer (probably AMI) arbitrary, or was it done to capitalize on some inherent advantage this offers?

RE