Folks, listen to this, the plot thickens... The chips ARE all working, and the crystals ARE working. How can I be so sure? Just listen to this... Now the breadboard assembly IS working, but the new PCB is not! I've programmed a very simple test program that just flashes the LEDs attached to the P1 port (thanks to someone for the tip), tested all three MCUs and several crystals, and the LEDs always flash happily on the breadboard. Now the only thing that can make the breadboard oscillator stop is to remove all caps from the power supply, but I believe that's expected already. And you were right, removing either or both 27pF caps from pins 4 and 5 does not make any difference.

I must say this is another breadboard, but it is identical to the first one. I don't think the breadboard is to blame, though.

Maybe it's a silly (and probably wrong) idea, but I suspect something else now. The first breadboard assembly used an external 78M05 power supply and it didn't oscillate. My new PCB uses another 7805 and does not work either. However my new breadboard assembly has a Sharp PQ05RD08 regulator instead of an external power supply and DOES work now! I've measured all three voltages and they are all the same, all about 4.97 V.

If all these assemblies are confusing, see this picture which I hope will help clarify things.

Next test: I'll disconnect the PQ05RD08 and go back to the original supply and see if this does change anything. I'll let you know.