Installing SP6a on NT4 SP4 workstations that have been in use for quite awhile crashes the OS on about 1-2% of the workstations. A co-worker said it happens because some of the critical operating system files are physically located outside a limit on the hard disk (2Gb?) and when they get udpated by SP6a it causes the OS to crash.

I can't find any documentation to substantiate that claim. I'd like to know if it's true and if there's any way to determine if this condition exists and if there's any way to correct it before upgrading from SP4 to SP6a.

There is a well know and well documented [until MS took it off their site last year - at least I can't find it anymore] of a bios/ntfs boot bug.

If you exceed 7.6gig for system AND partition [like you used Partition magic to create the partition] due to using int13 addressing the bios can't find the boot files if they are move beyond that point.

I really doubt its sp6 causing the problem. it has been a very stable sp compared to sp2 [that took entire corps down] and sp4 which was crap.

A little known workaround to this issue is to use a defrag utility like Diskeeper and mark the files boot.ini/ntldr/ntdetect/bootsect.dos as unmovable. This will keep the files at the beginning of the drive. It is usually defrag that causes this issue.

While on the subject have you defragged before doing the sp6 update? Very important to do and often overlooked.

Thanks, wanderer, for pointing me in the right direction. I did a Google on ntldr and 7.6 and got all the information I needed.
In this user's environment it probably isn't efficient to do a defrag before SP6a. They have about 14,000 NT4/SP4 PC's. Defrag before SP6a on 14,000 would take more time than rebuilding the 1-2% that fail and it would probably help to have those rebuilt.