Optimal Reduction

This is what Andras Kovacs has to say:

My view on optimal reduction (that I've mentioned a couple of times) is that it's only useful for higher-order programs, but in practical programs it's better to try to compile down to first-order code, than to try to make higher-order functions more efficient. Essential higher-order evaluation is very rare in practice.

Basically, optimal reduction offers improvements which look like fusion at runtime. But the alternative is fusion at compile time, and if that's ergonomic and strong enough, then we don't need the heavy runtime machinery