OK, so here's a pic of the controller PCB and its surroundings in my parts G81-3000 (that new-style case really is a bit of a pain to open btw):
Now I guess I'll have to remove it without screwing anything up.
OK, that's off. (Think it's Torx T8 screws.) Unscrew, then lift up at top left corner.
Hmm, that looks suspicious:
For some reason, there is something on the contacts, at least those with the traces coming from below. Some mechanical issue, e.g. traces rubbing when the case flexes or something?
OK, I'll grab my can of contact cleaner and some cotton swabs now...
On with the cleaning: Upon removing the goo on the traces, I discovered the potential root cause: Underneath there is a black plastic rim pressing the membrane against the PCB contacts, and for some reason this does not work equally for all of them. Maybe there can be some relative movement here, which abrades the silver traces on the affected contact sets. And no, I'm not an Athlon unlocking guru with conductive silver paint at hand.
Needless to say, there still are a large number of keys which register very intermittently or not at all.
IIRC, the G81-12100 has the old-style connectors for membrane interfacing. Guess they didn't want to spoil reliability on what was a very expensive board at the time...
Oh, and only having looked more closely at the first post now, the OP's G81-1800 seems to have something more like the old-style connectors. I'd try unlocking those, possibly cleaning the contacts and putting things together again.The OP's G81-1800's failure pattern would coincide with a problem for lines {n,j,m} in
huha's traced G80-3000 matrix.