You are right of course, that's why fuzzers are so good at detecting problems, but we do still need a proper test-case for the bug report. Explaining why it is silly, or indeed that silly is helpful, is already understood, but does not make it any less silly. It is silly because it is unlikely to have real-world parallels, which is why Richard made a test case that may actually appear in a real-world DB, which then becomes less silly but still demonstrates the problem to inform the bug report and possible future test-case. Just trying to explain it better because I do feel there might be some language barrier - like when you replied to Keith in argument, when he actually responded to Luuk informing him that his assumption is not correct, so in essence agreeing with you. I'm interested though, if this tool you use is not a standard fuzzer, but I assume produces random-built SQL sequences that should be workable normally. I assume you have a pre-built DB where you control what types of SQL sequences would make sense and allow them to be concatenated into test queries. If it is more elegant than that, I'm interested to know, if you are interested to share.