There have been a couple of threads recently on this topic, including my own query.
Out of curiosity I have dismantled a spare ‘octopus’ I happen to have, and think I now see how it works and why the light can come on.

Removing the electrical switch is indeed pointless and risky. It is not actually a switch, only a spring-loaded connection to the sliding piston. The latter has a rubber-sleaved central section, so when centralised the feed is not grounded. If a pressure differential occurs between the two circuits, the piston moves so that bare steel is exposed to the contact, the circuit is grounded and the light comes on.
The electrical switch is made of soft threaded plastic, and is a very tight fit. Even if it can be removed without damage, it is very difficult to re-start the threads.
Even when removed, it is still impossible to re-centralise the piston because there is nothing to gain leverage on
Therefore the only way to centralise the piston is by removing the end plugs and pushing with a thin rod, or by applying a differential pressure that reverses the pressure that originally caused the piston to move.
This gives the clue as to what causes the light to come on after work on the brakes, such as replacing calipers, or even just pads. When the pistons have been retracted, it is important to re-apply brake pressure very gently at first, giving the pistons on each circuit the chance to re-locate evenly. If something causes one piston to stick momentarily, and the other to go with a rush, this presumably causes a pressure differential which triggers the sensor.
My suggested way of rectifying the problem would therefore be to try and reverse the differential, by retracting the piston on the previously higher-pressure circuit, and then giving a sharp stab on the pedal to create a pressure surge on the other circuit.
I have not actually put this into practice because the problem appears to have cured itself.
Incidentally while experimenting and trying to re-centralise the piston it is better to disconnect the wire from the octopus and use a simple conductivity test to ground.
Finally I would not advise trying to dismantle the unit in situ. The end plugs retain lots of small spring-loaded components.

Hope this helps someone – it has me !