I thought it might be useful to share the resolution of this problem, given that no one could pinpoint it.
In frustration I took it to a local independent who used an interesting technique of pumping a visible gas into the engine (having agreed with me that it was not an electrical problem) to test for leaks. I was informed that historically this technique used propane (invisible) and a sniffer had to be used to detect leak locations - obviously imprecise. The colored smoke is much easier to 'see'.
This diagnosis disclosed leaks at the fuel injector seals (#7 was especially bad) and smaller leaks at the intake manifold. Having just replaced all the injectors and seals on my 560SEL it was frustrating to have to pay someone to do a job I can do competently, but I paid my $473.88 for the injector seal job and the problem is (largely) gone. SF Bay area pricing.
The cost of fixing the intake manifold leaks was quoted to me at $1,400 - evidently lots of labor hours are involved. Does this sound fair?
A related point. Earlier in this thread I wrote: "Followed the diagnosis in the MB 116.96 factory engine manual at 15.7 III-562/7. This states that pin 3 of the diagnosis socket (LH front wheel well) should return battery voltage against ground with the ignition switched on (I asume that means engine not running). I got only 3.80 volts. Next step is to test voltage between diagnosis socket pins 5 and 4 which should return 0 volts - I got 2.81 volts. Book says this means the ignition switching unit s/b replaced. Before I replace this expensive part, does any one have any input which might invalidate my diagnosis? " I checked my findings with Barry Stark (a 380SL owner and a good diagnostician) and he confirmed the MB manual appears to be incorrect. His car runs perfectly and returns similar voltages to mine.