I grab a png image using the radar api every 15 minutes and then combine these to create my own animated radar image for the previous 12 hours. I have noticed recently that a "rogue" image creeps in, sometimes quite frequently. Its always the same image (I will try and attach). I also do a similar process on my brothers website using a different api key, different location etc and likewise get a rogue image, although the image is different to mine but always the same that appears.
I have deleted all my png files (my script automatically deletes images older than 12 hours anyway) to make sure somehow a rogue image was not being used but to no avail.
If you take a look at my website http://midlifedad.me.uk/weather/radar.php you may be able to see a live occurrence.
Any help or pointers would be gratefully received.
Sorry to say that the same image re-appeared at 19:00 and 19:15 UK. I deleted all my png files when you stated it should be resolved but exactly the same image as the one originally posted has re-appeared. Please see http://www.midlifedad.me.uk/weather/radar.php
I have the same random behaviors using PiClock as I do in the browser, with no consistency in whether I get a proper radar layer or timestamp (UTC vs. EST). When did the fix get implemented, so I can monitor through out the day?
EDIT: I just looked at the screen, and the time stamp for the "region" is in UTC, while the "city" timestamp is in EST. The radar layers "seem" accurate and agree, although the UTC timestamp is indicative of a forthcoming "cached" image set coming.
I also see the same happening on my brothers system, this time at 03:00, 04:30, 07:15. Its always the same image, but different to mine. Here is that URL:
Hope it helps finding the issue.
I've been experiencing the same issue, within .Net I casted the byte to an image and checked the frame timings property to make sure they matched the delay I asked for in the query string, if it did not I had it wait a second and retry up to 5 times, this seemed to work until the issue is resolved.
The following seems to be pretty stable and returns correctly: