OBD1 DME generates non-specific fault code

[tgschneider]

I have 1995 E36 M3, S50 Engine, 10/95 build date, OBD1, 92K miles. Car runs great and passed smog a few years ago with excellent numbers. Then I had the dreaded DME flood experience. Had it fixed by ECU Doctors (EWS II delete version), cleaned the DME connector pins thoroughly (DeOxit) and checked continuity, and car again ran fine except CEL was always on, with fault code 200 (0xc8) which is "Bad DME". ECU Doctors says they got no faults during their test. After I reset the fault code (Peake reader) the fault comes back every time the engine runs. ECU Doctors will no longer deal with 1995 M3 DMEs so I can't get a retest.
As a test, I bought a used DME (same part number as mine) off EBay and put the ECU Doctors EWS II delete EPROM in it. Car ran fine except got same fault code behavior as above.

So, the problem occurs with ECU Doctors repaired DME and EBay DME. I think the problem is most likely not in the DME, but is related to my car specifically. The DME must be seeing something from the car that makes it generate this non-specific internal fault. Could be it checks its outputs as well as inputs for proper function, and logs the Bad DME fault (If only I had the source code for the DME software the source of the failure would be easy to find...)

I can't pass CA smog until I get that CEL to turn off.

THE SMOKING GUN: I stumbled across a starting sequence that be helpful in identifying the problem:
I reset the DME fault codes, crank the engine for less than 1 second and kill it (not letting it start). At that point, I can restart and kill the engine many times during the next 20 seconds or so without any faults. Then I get the Bad DME fault again.

Does anybody have an idea about how I can get to the source of the problem? And what might be the significance of the 20 seconds in the above test?

Thanks,
Tom Schneider

(I tested all DME power and ground connections, cleaned chassis ground points. Tested DME input and output signals with engine running - sensor inputs show normal voltages and scope waveforms. Output signals are good. Power voltage levels are good.)