Friend of me has this problem on her 2005 or 2006, I forget. So, I bought a failing combination meter on Ebay cheap and looked at the circuit around the TLE4278 LDO and watchdog for the Fujitsu processor. I was able to easily reproduce the problem by powering the combination meter from a 12V supply (GND pin 14 and +12V pin 21). If I interrupted the power supply for a short moment (slightly less than 1 sec) then the Fujitsu processor did not boot up quickly enough to reset the watchdog in time, so the board got into a funk of trying to reboot several times per second repeatedly until you either interrupted the power supply again to stop the watchdog longer on power-on and allowing the Fujitsu to start properly resetting the watchdog, or I could also get the board out of the funk by touching the Gate of the 12V power switching FET1 (marking J377) which is controlled by transistor TR3 which receives its control from the Fujitsu. Weird that, since the Gate is controlled by relatively low resistors R4 (15k) and R3 (10k to +12V). Anyway, the time-delay for the Watchdog is controlled by the capacitor C5 which is likely the minimum 47nF as suggested in the datasheet, which gives a watchdog trigger time of between 35 to 61ms. To check if my theory of speed of startup of the Fujitsu is correct, I added a 100nF capacitor in parallel to C5 (from TLE4278 pin 6 to ground) and after that I could not reproduce bringing the board in the "Funk" state where it repeatedly rebooted, apparently the Fujitsu is now fast enough to boot before the watchdog bites. NOTE: this experience suggests that the 12V ignition power to the Combination Meter has a short interruption while starting the car, this will allow the watchdog to reset the Fujitsu (due to undervoltage) but keep the watchdog circuit alive enough to immediately start biting as soon as the power returns, preventing the Fujitsu from starting fast enough to start triggering the watchdog. It seems that a simple and cheap 100nF cap soldered parallel to C5 and increasing the trigger timing likely be a factor 3 will take care of this problem and make the Combination Meter work normal again. Tomorrow I will install this modified Combination Meter and we will see how it behaves in daily operation!