I remember seeing something along the lines couple of years ago. Does the production portal by any chance have several server nodes or dialog instances as where other portals don't? If so, make sure to register item types on each node/dialog instance. In my case it was a portal cluster synchronization issue. Rebooting the portal cluster of course should fix the issue as well. Just to be sure, check defaultTrace as well as the backend system for short dumps at the time of registering item types.
↧