For tests with randomly generated inputs (e.g., FsCheck for F#), it would be great to have an NCrunch option that continues to run a single test until it fails, in order to catch rare edge cases.

Currently, the alternatives with NCrunch aren't so great:

1) If you set the maximum number of test runs to a large number, then when a failing case is found, it is necessary to run ALL the tests preceding the failing test in order to reproduce the failure, in order for the underlying random number generator to end up in the same state which generated the failure.

2) Running the test over and over again manually, waiting for a failure, is time consuming and tedious.