Possible, but with the pin you’re using I’d bet on power first. The Uno’s 3.3 V comes from a small LDO (an LP2985 on most boards) rated about 150 mA. An ESP8266 pulls 70-80 mA idle with Wi-Fi on and spikes to 300+ mA during transmit and association. The rail sags, the ESP browns out mid-connect and resets. When it does survive, every busy moment risks the same, which fits your half-hourly drop-offs.
What I’d do:
- Feed the ESP from its own regulator off the Uno’s 5 V: an AP2112K-3.3 (600 mA) or an AMS1117-3.3 module, with 10 µF plus 100 nF right at the ESP’s VCC pin. A 470 µF electrolytic nearby helps with the transmit spikes.
- Level-shift Uno TX to ESP RX with a 1k/2k divider. The ESP isn’t 5 V tolerant, whatever forum posts claim.
- Keep the Uno’s 5 V source decent. USB from a laptop is fine for the bench, a cheap phone charger often isn’t.
Michael’s point is worth checking after that. rst cause 4 is the watchdog, and a brownout that half-crashes the code can also end in a watchdog reset, so fix power and see what’s left.
For the Hermetry side, once it’s stable, keep the MQTT keepalive around 60 s. A short keepalive plus a flaky supply gives you exactly that pattern of timeouts in the Log tab.