STM32L4R5 not always booting after nRST

[RDowl.1]

We are seeing an issue with the STM32L4R5ZI that does not boot 100% of the time if nRST is asserted after powering up (i.e., pressing and releasing reset button). We have seen this issue with the Nucleo board and with two different boards we made ourselves.

We have seen this behavior using code generated by CubeMX 6.2.1 using STM32Cube_FW_L4_V1.17.0

Observations

• Failure is intermittent, it works for a while, then it fails for a while.  Could be as high 50% on average
• Once the L4R5 does not boot after reset, additional resets will not fix it; work around must be used

Work Arounds

• Power cycling always boot
• Flashing the board (St-Link) has a ~90% chance of booting
  • Note: A related problem is flashing the MCU only has a 90% chance of booting; flashing again or even a 3rd time will get it to boot. 

Has anyone else seen this, and is it a firmware issue?  

NB: I've used STM32 a lot and have not seen this issue before. Wondering if L4R5 SWD could be involved since flashing can recover it, mostly.

Robert Dowling

[Tesla DeLorean]

The L4+ will detect if the FLASH has a valid image, and with latch a BOOT-to-ROM mode that is only cleared with a power cycle. A blank or erased device will latch into this state.

Given the low voltages at which the core might be viable, I'd probably look to having a rather high thresholding POR circuit that holds the reset for a 100-200ms beyond that.

[waclawek.jan]

Maybe related to https://community.st.com/s/question/0D53W00000raYd2SAE/intermittent-break-at-address-0x1fff51f4-with-no-debug

JW