We're experiencing a strange issue using MQ tile images on our leaflet deployment (via opentripplanner). Sometimes, some images fail to load and are replaced with a gray block.
Upon further investigation, this seems to happen when a given image is unavailable from the server subdomain and instead of returning the tile image, it returns this along with a text/plain mime type:
"The tile you requested is not available at the moment. Please try again later."
Requesting the same image from a different server (otile2 instead of otile3, as an example) yields the image that was previously not available.
Should this be expected behavior? If so, is there a known workaround to fix besides detecting the failure and trying a different domain?
I also noticed that documentation (somewhere) mentions that each of the "otile" subdomains all point to the same CDN, yet if that were true this issue should not happen. Indeed, the DNS for each subdomain resolves to a different akamai server so perhaps at that level the image was stale? Shouldn't one level of CDN be able to refresh itself from another while minimizing things like this from surfacing and sometimes seriously affecting usability?
As an aside, the issue is aggravated by the way in which leaflet's TileLayer tries to parallelize the requests to workaround the well-known limitations of HTTP in that area, but falls quite flat when this issue hits.
Thanks for the help!