Continued investigation leads to the conclusion that the specific position one holds the board at results in the wire running parallel to a set of tracks. These tracks carry 3 communication signals from the wire to the IC. But the flow of current in the pcb tracks and consequently any electromagnetic field is opposite to that of the wire. Let me try some ascii art to illustrate
The top line and the vertical line is the wire. The middle continuous line is the pcb. The bottom continuous line is the pcb track to which wires are soldered. Imagine current flowing in the wire from left to right and onto the track. In the track due to our positioning current is flowing from right to left. Therefore the current set up a a magnetic field in one direction and on the track in the opposite. If we bring them close together - as we are - the two fields should cancel, thus cancelling the interfering current. Thus any unwanted electromagnetic interference gets cancelled and the circuit works.
The above is wholly speculative, but the only explanation I can think off. Given the length of wire and track this interference is in the 800Mhz and above band. The source of such interference is wholly unknown.
One method of eliminating such interference is to shield the wires. There are special wires made with an outer braid of wire or Al foil that acts as shield. Instead of again removing the wires and replacing with a shielded cable, I wrapped a piece of Al foil around the wires and grounded it by connecting an extra wire with a crocodile clip to ground. I was fairly sure that the circuit should now work. Nope. The circuit with cable shielded with Al foil now failed to operate in any position.
So, the malfunction was due to cross coupling interference between wires - signal on one wire interfering with a neighbouring wire. But why did this the cross coupling start after 3 years of use?
I decided to solder the signal wires directly to the IC pin, instead of their current closely spaced position.
First signal was STROBE. This signal enables or disables the keyboard. I soldered it directly to a convenient point near the IC, which is about 4 cm nearer than it earlier point.
VOILA. Everything works just fine.
I packed up the gadget and returned it to the kitchen where it is functioning normally for the last 2 hours.
But this repair almost had me believing in ghosts in the circuit ;-). While the problem does seem cross coupling, this is a design problem and should manifest itself at the design stage, or if it is just marginal, as intermittent malfunctions on the production line, resulting in higher reworking and/or rejections.
There is also this rare chance of some set of components in this specific piece have deteriorated just marginally to cause the problem. In which case we should have a repetition of the problem after a few weeks of operation…