Seeeduino LoRaWAN Vin supply behaves oddly with battery removed

Can anyone explain why Vin (as measured at pin 1 of J3) looks like the yellow trace in this shot? It’s a square wave between 0.4V and 3.6V at 0.92Hz.

From the schematics and the datasheet of the ETA6003, I would expect it to hold steady at 3.6V, until a battery is plugged in. I would expect the STAT pin that drives the CHG LED, and the BATT pin that tests for the presence of a battery, to be the ones that pulse instead.

Here’s the BATT pin overlaid on the Vin/SYS pin.

The issue I’m having is that the 3V3 rail drops out inconsistently every couple of hours, despite the 5V supply being steady. Given that the Vin rail is not steady, this could be the source of the problem.

Here’s one of the instances of the 3V3 rail (blue) dropping out. As you can see, there’s a strange interaction with the Vin rail (yellow).

Ooops! The yellow trace was actually CH_STATU which makes a lot more sense!

Still trying to get to the bottom of the 3V3 rail dropping out.