I'm a biologist so forgive me the poor wording, but there are "infinitely high" peaks/walls in the evolutionary landscape, as in: "the human urinary system has some evolutionary constraints and anatomical features that are not optimally designed from an engineering perspective, but they are generally functional and reflect our evolutionary history rather than deliberate design." <- this is from AI, I didn't remember the details, but the ureters or urethra take a long way around the inner part of the body because at some point it was a good route, but as we evolved it became suboptimal, but we can't just evolve to untangle all the pipes in our body and do some cable management... We're stuck with what we have (no tunneling electrons ;)).

So in that sense, a global maxima (minima?) solving function will probably not arrive at what we have, it has to take into account our evolutionary path, in its entirety (as in the whole tree.), because it is all encoded in 1 (or several) long molecules, and a radical changes would take a rewrite of many parts at the same time, which just does not happen. Maybe at some point we can do it ourselves... though I imagine there being a lot of unintended consequences in other places in the body :).

Another example is that sometimes complexity in a cell signalling cascade is simply there to get some timing right (ie delay some response), and the complexity is just there, and won't simplify by itself -> But then later the complexity may start to serve some purpose. Just like proteins that are "copypasta-ed" in the genome, leaving one copy free to mutate and attain a function, or not... etc etc.

Great question! It maybe can be, but what is the maxima that is being solved for exactly? There are usually tradeoffs, the most common and prevalent is our body's immune response. If it's under-active, we might die from an infection. If it's over-active, we get diabetes, lupus, chron's, and (probably) lots of mental health disorders.

So, it's a bit context dependent, and definitely dependent on other mutations that have occurred in other parts of your DNA. Most complex diseases are "polygenic", meaning it's a culmination of quite a few factors that would contribute to a specific good or bad outcome.

So, yes, it could be modeled as a sort very context dependent with a lot of highly correlated non-independent covariates. We do use quite a lot of statistical and ML methods to understand the genome (I work in statistical genetics), but the complexity of biology has so far proved a tough nut to crack.

Evolution is the ultimate "good enough". Once the required threshold is passed, pressure stops. The threshold might constantly be moving due to competition, environment etc.