I can confirm the shark fin like signals on my M5CoreS3. I am no I2C expert, but I think they might be marginal and so maybe just barely working on one M5CoreS3 but not on the other when talking to AXP2101. Maybe a stronger pull-up resistor might help?

BTW: I checked the I2C SCL line on an M5Core2 and it is much more rectangular, e.g. much more like I would expect it.

Thanks
Felix

The data seems to be random indeed, however, it is always the same.
We did notice shark fin like signals on the i2c lines on 400k. We reduced it to 100k but the results stayed the same. We do not believe that the bad signal quality is causing the current issue we are having. On a good working board the signals look identical.

For reference, here is inmage of the signal quality of the SCL/SDA line on 400k speed on the good board.:

The register values are taken with 100000 speed.
I think i forgot register 0x22. I've tested it and added it here:

We've also tested writing the value 0 to 10 on register 0x04. On the good board we can read back the correct value. But on the bad board it always returns 0x78

just curious, any particular reason you did not read out register 0x22? It is my understanding that this register determines whether the system powers off or does a reset when writing 0x01 to register 0x10.

Also curious, when the issue happens, is the battery full or close to be fully discharged?

The register values of the bad AXP2101 seem to be random. Have you tried to lower the I2C bus speed?

In my experience seemingly random issues often stem from bad / not sufficient power supply. Is the battery in good shape?

Thanks
Felix