I did something similar when investigating how SQL Server behaved under certain failure cases of the file system (specifically how it behaved when seeing lost w

I did something similar when investigating how SQL Server behaved under certain failure cases of the file system (specifically how it behaved when seeing lost writes). It's a really useful technique for validating behaviors under error conditions.