Coagie: Not so. If the inputs are screwed up, because a sensor is thrown out of calibration from heat or other damage, you could get a specific fault. Heat can cause added resistance, that makes 5 volts, which would be a "1" in binary, low enough, where it shows up as 0 volts, which is a "0". This could feed some radical numbers into the brains of your "complex electronic systems", and cause an intermittent, as the sensor heats up or cools, or consistent, but specific failure, , due to ambient temperature change. I think such a failure of some acceleration sensors was the cause of the Malaysian 777 stall in 2005.
Sensors are duplicated. The systems are duplicated (in the case of the 777 it appears multiple instances run in lock step with cross checking of outputs). Any differing values will be detected. So for your scenario to happen, you'd have to have all such sensors damaged in the same way and feeding exactly the same 'radical' numbers into the system.
The other thing that makes it seem very unlikely is the long time period (many hours) and small number of course changes with large intervals (based on the radar track at least).
Is it possible that systems damage caused the plane to change heading periodically but otherwise fly fine? Sure, but so are many other theories floated on here. Likely? I don't think so.
The 2005 incident you referenced was due to a software bug which caused the system not to reject a known faulty sensor reading. Left alone with no intervention, that one would have been catastrophic. Had that software bug not existed, the faulty sensor would have been ignored.