Beomaster 8000: Standby Issue - Beomaster Does Not Turn Back On During a 50 sec Window

Today I looked into a strange phenomenon I came across in the last Beomaster 8000 I restored. When no speakers are connected, or the speaker switches are in 'off' position, then the Beomaster would not turn on the output stages (i.e. no relay-click audible) if a turn-on occurs in a time window of 10-50 sec after turning it off.

This is not much of an issue in 'normal' operation, where one listens to some music, then turns it off, and comes back an hour later for more. But on the bench, when one tries out a few things and the Beomaster gets turned on and off in quick succession, and often without speakers attached, this phenomenon became quickly apparent.

This is the first Beomaster I encountered with this problem.
So I thought I would try to get to the bottom of this. The first thing is always to do some Googling these days. What came up was a Beoworld.org 'Vintage Products Forum' entry by Dillen/Martin where he states that:

 "It's not uncommon to see that a Beomaster 8000 won't switch the amplifiers back on right after being put into standby. It will light up the displays but not power on the amplifiers.
Put back into standby for 15-20 seconds and it will power up fine again. This is quite normal."

In the particular case discussed here the Beomaster does come on during the first ~10-15 sec, and then not until about 50 sec have passed. After 50 sec it starts up again. The phenomenon is completely reproducible, but the speakers must be disconnected/switched off to experience it. When the speakers are on, everything is fine.

Anyway, I opened it up again today and did a few measurements with my oscilloscope (since this is a transient phenomenon, we need a time dependent measurements...). I fairly quickly identified the cause for the phenomenon: Pretty much the only way to prevent the outputs from turning on in a reproducible fashion is via the fault circuit:

Both right and left channels have a 'fault output' that triggers the 'Fault Switch' on the power supply board to cause a shutdown of the outputs in case a DC voltage is present at the speaker output, or if the temperature in the heat sinks gets too high. In either case a non-zero voltage occurs at the voltage divider between 5R24 and 23 (on the left...that part of the circuit is on the #5 PCBs (output amps), which is used as 'fault output'

I measured the voltage at that R23/24 point depending on the time after shut down with speakers connected. The yellow trace in the scope shot below is the positive 54 V rail which was used to trigger the measurement at 48.8V, i.e. a short while after turning off the Beomaster as the rail started to decay. The blue trace is the voltage at the fault output. So we see that there is a ~600-700 mV negative voltage spike that decays within about 10 sec to zero. This is not enough to trigger the fault switch because of  6R62, and everything is o.k.

An entirely different picture presented itself after disconnecting the speaker from that channel:

Note the now larger 1V scale for the blue trace, i.e. the initially still present small negative voltage spike seen with speakers connected in the scope shot above was completely overwhelmed by a ~4V positive potential when the speaker was disconnected. This potential boost was reset after about 45 sec and zero voltage remained at the fault switch at that point. This corresponds nicely with the ~50sec I measured that I had to wait to get the Beomaster to turn on again. It seems that after 50sec everything resets. This correlates with the delay when the microprocessor turns off the 15 V rails after shut down has been initiated.

It may be that there is a potential build up due to an inconsistency in the decays of the positive and negative 54V rails after turning the Beomaster off. This is no problem as long as there is a speaker connected, which provides a path to ground if there is some DC transient on the output during shutdown. Without speaker, the potential builds up, and the fault switch is triggered.

The real question is: Why do only some Beomasters exhibit this phenomenon, while others do not. I am aiming to do similar measurements on one that does not show this phenomenon, maybe we can learn something from that...