I sent my ignition modules to Carmo in the Netherlands.
It took about a week to get them back (to the UK). Plugged them in and the 912 fires up perfectly. In fact, it seems to fire up better than it has done for a long time (never realised the start-up spark had got so weak). They cut out the back, repair 'something' and then pot the module from the back, thus making it completely sealed now.
I have been working with another guy in the UK here to try to figure this out (along with Kevin's very useful information). There is a heat dependency and clearly some component is 'on the edge' of its design margin (which is temporarily repaired by application of heat) - we have speculated it could be NPN hfe sensitivity or one of the SCRs having a weak gate threshold (but other evidence contradicts this at times).
We got to the point where it needs quite a lot of effort to really get to the bottom of (certainly doesn't seem simple on the face of it), in which case spare time return on investment is low! Keeping our noses to the ground if any more gem information arrives.