How should a background process behave if it encounters an i/o problem?

My customer's 10.2.0.4 database crashed last week. On examination, it became clear that there were some i/o issues, and the MMNL process had tried, and failed, to access a file. It was holding an enqueue at the time. It seemed to hang for around 4 hours until another process, which wanted the enqueue which MMNL was holding, killed MMNL, thus bringing the instance down. I have seen notes which explain that a process, if it has to wait too long (whatever that means) for a row cache enqueue lock, may kill the holding (blocking) process - and that this behaviour