Analogies can often confuse as much as help but I'll try anyway:
You have it pretty darned right -- wheelie away!
Look at it another way:
When a modestly smart human is given a challenge of making a machine better they apply what they know and think about the problem. They may sketch some thoughts and do some calculations. They ask more about why the current version is not as good as desired or what has changed to need a new version.
Only then to they pick what might be changed. A lever arm longer somewhere, more horsepower in a motor or more insulation (longer hair?).
Nature produces designs in an utterly different way. The way some of us dummies often try to fix something broken when we have no idea how it works. We tinker. We change this and then that and hope it works.
Nature does this to an extreme. It has no idea there is anything to fix. It has no idea of what is needed. It doesn't "intend" to change anything.
It just produces a slightly different version of the machine every time it is manufactured. Each human has a few small changes so there are 6 billion different versions running around. How many new versions of "Acme Cockroach" is produced each year????
Then all the bizillions of "fixes" that don't work out are thrown away. Any that work ok or better are kept. Then the whole process is done again the next generation.
Imagine if humans designed this way. Research labs would be factories that produced millions of machines a year each with something changed a lot or a little. Out back would be a huge mountain of things that were just too dumb to be shipped and out the front would go those that worked. The next round would start from only those that worked and more changes would be made.
By the way animals are "thrown away" in huge numbers. About half of all human fertilizations are thrown away (usually before the mother is aware of it) because they are just too messed up to develop very far at all and spontaneously abort in the first days. I'm guessing this is true for most animals.
It is hugely wasteful in material and a company trying this by actually building things to try them out this way would go bankrupt fast. But the designs can "explore" all sorts of different "fixes" and find many, many that work well and then better.
The answer to a "problem" in nature when "asked" what shall we try is:
e v e r y t h i n g.
(The above leaves out a lot which can be discussed later. Evolution is more constrained than I am letting on.)
™ Version 4.2