I seem to be experiencing the same problem. Was there any further development on this or did it just stop at the admin_init patch? It’s a great work around, but I have this irrational need to know what caused the problem and some version of a “Best Practices” solution.