It wasn't SQL, no. I've actually got a bunch of mocked WMI objects I cobbled together, some AD objects, and a couple of others. I do it whenever I need a similar-seeming object (e.g., specific properties, most times) I want to make sure my code is working with correctly. Sometimes, I'll just have the mock run the "real" command (if it's non-destructive, like a Get-), and the serialize/deserialize those objects to create "dead" objects with the same basic members.
Ok makes sense. What I'm trying to avoid is having tests that always pass without the mock objects actually representing the real objects, for example when a DB schema gets updated. By having a single version of each mocked object in a central file, will be easier to update in one place and then see how all tests handle it.
Seems more robust than having multiple versions of the mock objects anyway.