A basic ESP12S and a small BME280 breakout. Nothing fancy, no regulator, just some decoupling caps. Powered direct from the two (in series) AA batteries.
Data sent by UDP broadcast packets.

Previously I had been using 1-Wire sensors but they have disadvantages.
A full resolution temperature reading takes 750ms and the data-sheet says 3~5.5V.
The BME280 produces a temperature reading in just a few ms and it’s working voltage is 1.71~3.6V so powering from two alkaline batteries is no problem. The ESP will fail well before the sensor gives up. The ESP should be good down to 2.5V (in theory).

With readings taken every 5-min, the lifespan so far is stretched out to 14 months which is more than I expected.

I currently have two of these running.
Even though they started out with near identical voltages, one has slightly higher drain than the other.
The yellow trace device is further away from the access point so maybe that makes the difference, or it may just be component leakage differences causing the slightly higher parasitic power drain.
The data break was when I messed up my logging configuration and didn’t notice for ages!

When the tests end I will hook then up to my Nordic PPK2 power profiler and test the leakage/consumption for both units to see how different they are.