Hi Mixa64, thank you for the suggestions.
I've measured a substrate voltage of -2.65V (with a Vcc = 5.06V directly measured on the CPU pins) using the small metal point you said:
About the INIT test, I've done as you said:
Power applied (1, 20 to GND and 40 to Vcc);
CLCI (16) to GND;
DCLO (26) and ACLO (25) connected together and with a 2k7 pullup to Vcc;
INIT (27) with a 2k7 pullup to Vcc (in parallel with a 4k7 pullup resistor to Vcc on the PCB);
but INIT is always fixed at Vcc, even when both pin 25 and 26 are to GND.
Also switching CLCI do not make follow on the INIT output pin (27) the value of pin 25 and 26 (connected together).
Yes, that photo misses a wire to Vcc I added later.
There are some pullups on the red board:
WRQ (10) and INIT (27) have a 4k7 pullup to Vcc on the PCB;
SACK (13), DMR (12), VIRQ (28), HALT (29), EVNT (30) have 10k pullups to Vcc on the PCB;
WAKI (11) is connected to GND on the PCB (with a jumper);
AR (23) is connected to SYNC (21) on the PCB (with a jumper).
I've checked their levels and are at Vcc (with the only exception of WAKI of course), so those pullups are fine.
The only pin that is working is CLCO that outputs the CLCI input clock divided by 2. All the other pins seem fixed whatever you do.