You won't believe how many people that actually use NOLOCK or READ_UNCOMMITTED to "fix" concurrency and locking problems. One of our vendors use NOLOCK in every(!) SELECT statement in a system we have purchased from them. They have never bothered to explain the reason for doing it, but my guess, after doing some investigation, is that it is caused by two things:

- Lots of missing indexes, which results in excessive scans

- Bad programming, because when I save data in the front-end, SQL Profiler showed that before the transaction for the update statement was committed, another connection from my pc was opened trying to read the updated row(s).