Should this be a standard for loop or doesn't matter?

It could be. I've been taking an approach where I use ES6/fluent code unless something is provably a performance bottleneck.

I'm with @alfaproject on this, may as well write a for loop. One less function allocation per call to this function.

I find the fluent form much more readable, and it's used throughout the compiler today. I would rather not optimize prematurely. If benchmarks show that these temporary allocations (which are short lived anyway) cause significant performance issues, then we should optimize it. Until then, my goal is to keep the code as readable as possible.

I meant:

for (const value of values) {
  const [field, property] = value.split(',').map(piece => piece.trim());
  map[field] = property || field;
}

which is just as readable (I think) with the benefit of being more performant.

You might considier making this into a pure function that reduces into the map. It can be used to create the map above.

Done.

remove the map() here, and do the trim() on the next line where you need to. One less Array allocation.

As we discussed, I'm not worried about optimizing the allocations at the moment.

I'm with @alfaproject on this, may as well write a for loop. One less function allocation per call to this function.

I know you didn't touch this, but is the omission of a hasOwnProperty check here intentional?

And as below, this could be a for..of loop, that will save a function allocation per turn through this loop.

The same questions/problems exist in other code below (that you didn't touch) in extractHostBindings

I know you didn't touch this, but is the omission of a hasOwnProperty check here intentional.

Can you elaborate on what the hasOwnProperty check would check? I don't follow.