When you optimise a program, you do so by looking at how things work and then optimise for that specific case (e.g. "instruction A runs faster than instruction B and can achieve the same result, so let's use instruction A everywhere"). This comes at the expense of lower performance in hypothetical worlds with different laws of physics.

That ""500k a year noncompete" is a great hypothetical, but it doesn't happen in this world, so we shouldn't worry too much about sacrificing it to improve performance of a hotspot.