There were a couple of updates today to the delivery code. Please see the notification thread for more details.
Workaround: It has become apparent that numerous images were not properly migrated during our storage hierarchy move late last year. Since we have needed to keep "rolling refresh" (keeps refreshing images monthly) OFF, those never got re-captured. So, I have decided to add back a file_exists check to detect the missing images as they are requested. At that time, the request will be automatically put back into queue for re-capture, free of charge. Soon, we will eliminate the reason keeping us from running the "rolling refresh" and this won't be an issue any longer. Until then, this should alleviate the burden from customers.
BUG FIX: The Updates for 02/26/2011 introduced a couple of issues. None were reported but were discovered during troubleshooting of the missing image issue (above).
Details: The requests were being double decoded in some cases and stored in the database incorrectly. This caused unnecessary delays and odd delivery behavior.
Solution: Today's changes should better account for encoding and will correct the double decoding problem. We will most likely delete the broken requests and let them re-queue naturally as people request them.
Enhancement: I will be adding a little bit of code to modify the tick mark (') detection that was added during Updates for 02/26/2011. Currently, it allows for a malformed request that wastes capture generator resources. Once corrected, it will correctly handle the request with an "Invalid Domain" response.
Note: We were also informed that before the changes were made on 02/26/2011, it was possible to send through the malformed request and get back a valid screenshot. Since the malformed request is not "best practices", the changes will no longer allow them.