The FAA has just sent this one around:
Fairly obviously somebody wired up some lights without the two wires being tightly close together, so a magnetic loop was formed, and one would expect that to screw up the fluxgate magnetometer compass. That is for example why one never returns wingtip light current via the wing metal; it always comes back via a wire.
But why would it affect the pitch and roll?
Cropdusting at night might have a massive work light. Surpised an autopilot was in use.
The question I have here is why they never checked this on installation. I suppose this worklight was present then?
Other than that, I agree, this is probably the result of a rather specific installation. Our low power lights would hardly be drawing enough current to do something like this, particularly if they are LED.
The problem in this case is the effect the mag field (assuming it was that) has on the pitch and roll. That is really unexpected.
I wonder if the pitch and roll indications remained erroneous the entire time the worklights were on, or whether the pitch and roll indications were only temporary (immediately after switch on, and immediately after switch off only, rather than constant).