So, since I started this thread, here is some update on the current status. The good news first is that I did not break any boards since this discussion started 
, but of course, I faced problems from all directions. One of the biggest issues was a loose magnet, which turned out to have a very strange effect ( we discussed that in a different thread). I only noticed this, becaue I modified my B-G431B-ESC boards to support a full SPI and now I can monitor the diagnostic outputs of the sensor and when the magnetic field deviates too much from the expectations, then I stop the motors. In addition, I now have sensor calibration and use the same routines but much coarser for a quick sensor test at startup. These steps eliminated one source for errors that might have killed my ESC boards in the past. On the BEMF side I reduced the supply voltage to 12V and added code to closely monitor the voltage on each board. The results were surprising in that the generated EMF is much higher than I expected. I measured up to 16V for a 12V PSU and that was not even close to an extreme operation of my device. Adding 470uF per ESC did not change much, that’s simply to little in this case. I will go for 4700uF, but need to order capacitors first. Also, I implemented a circuit similar to @Valentine 's, which solved the overvoltage issue, but now I have a problem with undervoltage instead, as soon as the protection circuit kicks in 
. Hopefully the higher bulk capacitance will solve that. I will also go for a higher supply voltage again, as soon as I manage to get my hands on a 20V/5W zener diode, but these are hard to get in small quantities and at reasonable prices these days. Apart from that I fought with my CAN bus, redesigned the mechanics of my paragliding simulator, use torque instead of position mode to control the break lines and all this together makes me very happy since operation is very smooth now and feels quite real - as long as I switch off the over/under voltage protection (which I only do under very controlled conditions and at 12V now, I learned some lessons…).
As an advice for other users of the B-G431B-ESC1: Be very carefull with your power supply, take generated EMF into account, do not, really never ever, exceed the max specified supply voltage, not even by a little and not even for a very short time! Use the boards option to monitor VBus and if you have the chance implement sensor diagnostics, then do so! The B-G431B-ESC1 is really good, but requires some love, good soldering skills and the on-board aux power supply is very, very sensitive to overvoltage.
Happy holidays
Chris