This may be related to bug 855721 which is a problem in the way that NT handles semaphores during the mount process. This evidently doesn't occur all the time, but can be reproduced as follows:

1. Mount database failed (due to controlfile corrupted).
2. Restore controlfile.
3. Mount database again without shutdown -> Problem occured!
When tried to mount the database, got ORA-1102(cannot
mount database in EXCLUSIVE mode) and found following error in alert.log.
Once this error occured, they can not mount the database even though they
shutdowned the database.
.
ORA-09341: scumnt: unable to mount database
OSD-04400: unable to acquire internal semaphore for process
O/S-Error: (OS 183) Cannot create a file when that file already exists.

This error occured only when they tried to mount database twice. And
first try failed with error. For example..
.
1. Mount database failed (due to controlfile corrupted).
2. Restore controlfile.
3. Mount database again without shutdown -> Problem occured!
.
oracle80.exe create a semaphore scumntsem_ at mount. But this
semaphore is not released properly when mount failed. I guess the handle
of semaphore is lost at above situation and oracle80.exe can not release
it correctly. To mount the database, the service of oracle database have
to be restarted.
This problem don't occur on solaris, so this is port specific problem.
And I confirmed this problem occurs on all release of Oracle8 for
WindowsNT.

Solution :To restart service.

1. Start Service.
2. Startup nomount.
2. Rename controlfile to another name.
3. Mount database(alter database mount) -> ORA-205 occurs.
4. Rename controlfile to original name.
5. Mount database(alter database mount) -> ORA-1102 occurs(reproduced!).