I have seen similar issues.

The long startup could be because the database was doing automatic recovery after dealing you the crushing blow. If you had a long running transaction with lots of changes but no commit, it can take just as long for the undo/recovery.

Since it was in recovery, you ran into the io overlap issue.

You should check the alert log to see if it was in recovery (many of the later releases of Oracle will but a statement about recovery starting and completing there.)