One thing we can improve on is the number of reads of the log data on each ‘view log’ request. Currently every time you click on ‘view log’ it makes 3 requests, one for the regular log data, one for watering level data (wl) and one for flow data (fl). This was done to make it easy for the UI to calculate average watering level and so on. But the downside is that reading the log three times in a roll incurs a lot of overhead, and I suspect it’s also relevant to the rebooting issue reported.
While I agree there may be some quality control issue with ESP8266, I think its reliability can be improved in software. Also, the fact that it’s extremely inexpensive make it easy to replace if any problem occurs. OS3.0 uses separate top-bottom layer design, so if the top layer (containing ESP8266) has any problem, it’s pretty easy to toss it out and replace it.