In your spare sonoff you could run Tasmota to see if more stable at same time.
The captive Portal is indeed a little heavy on the esp82xx because it internally needs probably the web server component to work 
Please note that enabling this component will take up a lot of memory and may decrease stability, especially on ESP8266.
Web Server Component — ESPHome
To get a little more in-depth about memory usage (specially now that you have the 2 devices behaving differently) you may use the debug sensor which will inform about heap memory of your esps!
And is your yaml very complex/with a lot’s of compoents anyway or is it rather a simple one?
Both devices have been running with 100% uptime since disabling the fallback AP and captive portal. Even during periods of poor wifi.
So it really does look like the root cause was that the captive portal is too much for the ESP8285 chip in these Sonoff devices.
Rather simple. Basically the following yaml with a couple of scheduled automations (time:) and the captive portal.
2 Likes
thx for the infos.
I have a similar issue with an ESP32 and weak wifi.
It reboots every ~4h.
Did you let the “reboot_timeout: 0s” on api & wifi set?
Now i am trying to deactivate: fallback AP, capacitive portal, webserver & reboot timeout
I know the ESP32 is more powerfull, but i dont know what todo instead of this.
Yes. I set reboot_timeout to 0 but it didnt help.
Enable debug: to see the restart reason and determine if the restarts are expected or unexpected.
thx for the hint,
i have now activated the debug component like described here:
I hope there is some usefull information in the reset reason after the issue occurred, and not just “Software Reset CPU” 
There is just an PowerOnReset in the debug Logs.
Also i was connected ofer wifi to the device with live logs and NO exception or anything. Log just stops!
Here is a possible workaround (set static IP instead of using DHCP): Help with esphome error/rebooting every 2-3 hours - #4 by pOpY
Will post updates in the other thread.