The Hobo datalogger buried in the dirt under the patio kvetched about a low battery, which produced this surprising result:

Cells from the same lot have been doing just fine in the other dataloggers, so I hope this is a one-off weak cell and not the harbinger of another run of dead cells.
Alternatively, there could have been a leakage path.
I’d believe leakage over six months, but not two!
On the other paw, it’s buried in damp dirt, the O-ring may not make a perfect seal, and the tiny silica gel packet has long since lost its mojo. Sooooo, ya never know.
I should gimmick up a tiny constant load and voltage monitor to test coin cells; the CBA lacks any accuracy below a few tens of mA, because that’s not what it’s intended for.