Personally I tend to only grade the last bubble and I often look for some specific thing (like good ID Names or format of the code). I piloted last year before the automated checking was done and I still don’t really “trust” it. Sometimes it seems like it lets people go with things I don’t think should pass and others it seems overly picky and frustrates everyone.
I also like to emphasize creative components so I tend to grade the mini-projects where there is some room for self expression.