@David-Ryzec-3 I placed the timer code just before the HTTP request. Looking at the log, there haven't been any errors in the past 7 hours, so maybe it's fixed, but it could just be luck. We get data every minute, but I'm not in control of that and can not let Windy drive the sample rate. I can of course change the code to send to Windy only every 6 minutes instead of 5 and that will eliminate the issue. I'll switch to a 6 minute send period, if the problem returns.

@Vladislav-Iaroshchuk I don't care all that much about reporting on Windy (it's kind of our third option for viewing the weather station data) and the spec leaves things open to interpretation.

Now, let's say you are driving in a car and you have a speedometer. Should that dial show your sliding distance traveled in the past hour or should it show your speed at the moment you read it? That's why I think a sliding one hour average of rainfall is less than helpful.

What's worse, unless you can go indefinitely far back in history to a point where the average rainfall was zero, there's no way for you to derive the instantaneous rainfall from an hourly sliding average.

If I'm looking at the data as a human, I want to see at a glance if the rain seems to have ended. A declining average might tell you that it has ended, but it might also mean that it's raining less. A per minute graph is more helpful.

I'm getting relatively frequent errors when I try to send data every 5 minutes. I could extend the period to 6 minutes to fix the issue, but I think it's weird that you say that 5 minutes is OK, but then even a slight amount of jitter in the period will cause error messages.

"Measurement sent too soon, update interval is 5 minutes"

Here are some samples of the measured interval in seconds that show the the variations are around 60 milliseconds and we're still getting errors:

300.001
300.037
299.953
300.035
299.992
300.036
299.943

As an experiment, I'll add a static delay before sending the data. It will not affect the interval in any way, but it might help if the problem is in fact a clock sync issue. I could of course track the interval and add a variable delay, but our station is on a strict one minute clock, so the Windy data would start to drift relative to that clock since there's no way to claw back the added milliseconds.