Outline of the problem: The Model M consists of two membrane sheets and series of buckling spring actuators located over them at various points. On one membrane sheet, you have traces running from the controller, on the other, you have ones running to the controller. They are laid out in such a way that when one sheet is placed over the other, each trace on one sheet intersects with each trace on the other sheet at exactly one place, which corresponds to where the keys are. At rest, a third blank sheet with holes for where the intersection points are located separates the traces. When a key is pressed, the membranes are slammed together at that point. Current flows from the controller, through one of the traces, through the trace it intersects with and back into the controller at a different point. The controller is able to determine which row intersected with which column, and thus identify the pressed key.
Membranes don't last long, and over time, the exposed traces can wear down, or the contacts that plug into the controller erode (when you open it up, you'll see a little flat cable coming out of the inner assembly which plugs into the controller), and the keyboard controller stops picking up keys. From experience with a Model M where the Alt keys died, the two Alts are on the same trace (on the lower sheet I think) so the odds are that if one go, the other goes.
Since the trace on a membrane is just a strip of conductive material, you can fix a fried trace with a conductive gel pen (I used one of these
). Make a list of all the dead keys on the keyboard, and check out the membrane sheets. You'll more than likely see that most of the dead keys will correspond to one or two traces on one of the sheets. Using a multimeter, test from the start of that trace to the end - it will show nothing if the trace is fried. If the trace seems to be alright, check to see if you have conductivity between the end of the trace, and the contact for that trace on the ribbon cable. Experience tells me that the contacts fry quite easily compared to the inner traces themselves, usually fried contacts will be completely black. One test that you can do from the outset if you see suspicious looking contacts is to position one probe of your multimeter at one end of the contact, and the other on the other and see if you get conductivity.
If you are ridiculously lucky, sometimes there is a bit of exposed trace above the contact that you can strech out with the gel pen to form a clean new contact, and you never have to access the membrane. Unfortunately, I've only ever seen this on later Model M types, so in your case you actually have to access the membrane in order to 'draw' a new trace on top of the flat cable from the membrane sheet to where the old contact is.
Getting to the membrane is where the real fun happens. The Model M's internal assembly consists of the plastic top where the keys slot in, secured to the steel backplate using plastic rivets, and the membrane is sandwiched in between. If you want to get at the membrane, you need to hack all fifty-something rivets off with a chisel, there's no way to non-destructively open the assembly like there used to be on older IBM designs. Now, when you go to put everything back together, you need something to hold the front to the backplate... glue has been tried but doesn't work well, so the only thing for is to get out a drill, drill holes in the plastic upper where the rivets used to be, and used screws and nuts to secure the thing together.
On the up side, a lot of long term durability issues involving Model Ms are caused by those rivets popping, so when you replace them with screws and nuts, you get a nice consistent key feel, and a more clean clicking sound (when the rivets start failing, Model Ms start pinging a lot)
There's a detailed guide on the rivet removal/replacement process here.
Apologies for the tl;dr post Edited by ch_123 - 1/23/11 at 1:55pm