Once the document chaining effort is done, we should write a test that spams analyze and verifies that the samples collection is valid (i.e. there are no "partial" samples) repeatedly in another thread (or something like that) now that on analyze we have to delete more than one document from the old sample.