Checking the sources in kernel, I have reason to believe that the issue should be fixed in software in the latest OS2006 release. At some point in display init code there is a comment about going in wrong sequence would physically destroy the hardware...