Btw the problem seems to have gotten worse for me. I haven’t done full analytics yet, but it seems like I’ve got a 5% false alert rate, roughly 1/2 high-flow and 1/2 low-flow. If I was the person writing the software (and I am a software engineer and am familiar with issues with time-series monitoring data), I’d alert off of a moving average requiring a minimum number of samples. I suspect they are getting a pair of bad readings (high followed by low or low followed by high) and taking it as gospel, which is a classic formula for causing false alerts.
Also go home, Rachio, you’re drunk:
