I dont know if it was a network problem on February or something else, but it is probable that the image was (auto)downloaded during those 22 seconds while ACP was "not doing anything". The weather station polls take normally less than a second and it pulls the data from network only every 5 minutes. The network is down very rarely.

Of course the network pull should timeout sooner, I will fix it. But unsafe condition would have ended the observing script too. Maybe it should have, but the weather station data is not so time critical, it gives just the humidity and wind speed and both are usually changing slowly. I could take my chances observing rest of the night without the weather station data if the beginning of the night was safe. The script now sets the weather "unknown" but not "unsafe" if/when the network data can not be retrieved. I did not know the connect has so long timeout.

So the weather poll IS blocking the camera poll? Anything else in ACP?