Blood types aren't fixed across the population. It's kind of a tricky question which of our points they make a better example for. :D

But also, I guess by "feature" I tend to mean "noticeable feature". You might be more right about that, at least as far as terminology goes. At the bottom end you have a point mutation to a codon for the same amino acid, and then eventually you get something like a slightly different color pattern. Those are probably free.

But as soon as you have to, say, add or increase the size of a physical feature, there's a metabolic cost. And when those features aren't used, that (small!) cost drives them to disappear in relatively short order (for instance, all the convergently legless lizards). That strongly implies that for those, there needs to be a persistent reason for them to keep existing.

Entirely aside, I wonder how long we need to have blood transfusions as part of society before blood type compatibility starts exerting selective pressure.