Interesting, you're tempting me to pickup a couple M250s as spares just in case the worst happens to Enphase, but I'll probably hold off for now. I think if worst came to worst, I could also use the S280s using the same cabling and Envoy-S, or am I misremembering?
Line 207 - all of my 12 M250s are still in good shape. Have now reached 6.3 MWh total production.
I've seen the brief return of the inconsistent consumption metering, but it's become a very rare occurrence. I do think it's some combination of highly volatile solar production (peek-a-boo sun) + the Envoy-S sometimes bogging and not comparing production meter vs consumption meters simultaneously.
The issue is never detectable on a steady sun day, but doesn't always happen on a peek-a-boo sun day either. I'm guessing the Envoy-S bogs sometimes on steady sun days, but since the output is steady, the non-simultaneous readings don't produce a noticeable enough misreading to be casually detectable. (Unlike on peek-a-boo days, where a split-second offset can result in reported very low or negative consumption levels.)
As an side, it seems Enphase has started making Envoys only report 15-minute intervals, down from 5-minute intervals, as a cost saving measure. This results in stair-stepping on the Enphase graphs (and anything that pulls from their API). Very annoying. In order to get 5-minute data back I've had to switch pvoutput from API polling to pushes from my local server polling the Envoy-S. Seems silly. If they want to cost save, they should just reduce the >30 day 5-minute interval data down to 15-minutes. Tiny savings either way, but the storage cost of the long-term data is going to way outsize the last couple weeks of data.
edit: typos