LTC4421 start-up issue

Moreover, if I try to shunt IntVcc and SHDN (through the switch) and I only have the battery connected, the system won't start and the FAULT2 pin is driven low. However, if I try the same on another PCBA (identical to the first one), it works fine. If I use another battery on the "faulty" PCBA, it works well. I don't think it's a battery problem since it is not consistent across multiple PCBA and because the battery connector is directly linked to the cells, so there's no way for the BMS to shut it down (and I checked the battery on a testbench, all good. Also probing the battery leads on the PCBA I can see that the correct voltage is present) 

Update: now it always happen with the same battery (I will call it BAT_FAULT in the following). With 5 different batteries it works perfectly. Again, if I use BAT_FAULT in another PCBA, it works. Difference between the PCBAs is the length of the cable that connects SHDN and IntVcc (passing through the red switch).

If I don't connect the load after the LTC4421's MOSFETs. everything goes well. If I connect it and try to power up the system, the FAULT2 pin is driven down. I'm starting to think that it might be a battery issue, but I don't really see how the battery could explain this behavior.

Update: I connected 200kohm pull-up resistors between IntVcc and:

- neg_VALID2;

- neg_VALID1;

Now, the system works with every battery, even with the faulty one. How is this possible?

I also soldered a 10 Ohm - 10uF snubber and VBAT input before soldering the above mentioned pull-up resistors, but this didn't change anything. Just to say that I also performed this modification.

Update: the system stopped working again. We tried to put a big capacitor (470uF) on the battery solder pads and now it works perfectly, in every condition. Apparently, the big oscillations of VBAT during power on were triggering a FAULT on the battery line. Why's that? Shouldn't the fault be triggered only during an overcurrent event? Note that the neg_VALID2 has always been 0V.

It's because prioritizors slam on hard into whatever your current limit is.
If you have some inductance, VIN will drop below your UV-OV window.
Instead of adding C_IN which defeats inrush control, add a small cap to the UV pin or lower the current limit.

Ashapiro I have no current limit, SENSE and OUT pins are shorted together and connected to the output voltage, that why I find it so weird. FAULT should be triggered only during an overcurrent event right?

Maybe some noise coupling to the sense pins during switchover.
Are the pins shorted at the chip, or further away at the sense resistor?

They're shorted at the chip:

But it might be noise induced indeed. I believe our current problem is mainly related to the prototype integration: many PCBs, long wires,... we will soon receive the final PCB, we'll see after the testing if the new version is immune to this issue.

Thanks a lot for your help!