Found the problem. First, there WAS a relationship between the failed keys. Below a certain portion of the keyboard, it was every eighth halftone that had failed. It turns out the keyboard is scanned in groups of eight keys, and that one trace which went toward the low octaves from the connector to the computer had developed a crack for no apparent reason (the board under it was OK). The trace that went toward higher octaves was intact. Soldering a jumper between the appropriate connector pin solder pad and the nearest diode on the trace route past the break solved the problem with all keys.

By the way, there are two contacts under each keypad. The difference in height of the rubber pads mentioned above is what gives the keyboard the capability of measuring velocity. The first pad makes contact, then the second pad. The time difference between them determines velocity. The note sounds when the second pad makes contact, and continues until the first pad loses contact. All contacts are make/break -- no variable resistance or anything fancy. The circuit board contacts are coated with something black (carbon? rubber?) to make them compatible in the long run with the rubber contact pads.

Hope this helps someone else.

Steve