I wonder if it is because neovim's testing is setup differently and they wanted to port the patch immediately rather than spend time which they might not have had at the moment to rewrite the test. I suppose the patch was tested upstream, though you would want the test in neovim to prevent a future regression.