On 09/04/2008, Jesse Keating <jkeating at redhat.com> wrote: > How does one "play safe" here? We can't reproduce it, how can we > possibly fix it, or even verify that it's fixed? And how long do we sit > on the possible fix (if we could ever create one) before we decide it's > "safe" ? You play safe by unmounting *everything* and doing a sync() ;o)