And in that case, the implementation of the solution will probably be highly
coupled to the specific features of the language.

you have the time AND the data to figure out the real bottlenecks of an
algorithm.

Because, in the real world, you will in most cases have the help of a VM that
does the optimization for you, or:

than a hard to decipher blurb of lines which can take a while to understand for
other developers.

I'll always prefer the most straight forward, easy to read and understand
solution for any given problem

I either know the optimal solution already or I'll Google it. What's the point
in that?

But what does implementing a performant solution in withing a short period of
time proof?

It's unquestionable that I should be able so solve an algorithmic problem
correctly.