Initially I more or less assumed it was due to some changes or problems at the service provider, and these were explored with some persistence. They "cleared the app pool" several times to clear the block, but it always came back.

I think the problem may have been due to a method in the hit-counting which inadvertently was not explicitly closing down a database connection each time it was called, and connections were being left floating in Limbo. I also found a rogue hacking program masquerading as a web page in the large Resources data store which, quite independently, looks as if it could have been causing problems. It has been removed. Some code has been modified to help monitor this possibility.

Graham