Coming back a bit late. Thank you for the detailed explanation. Now that the relevant file is found, things become much simpler to understand.

Side note (no action needed):

I noticed that, due to the in operators (lines 16 and 29), the code would be erroneous if looking for CSS property names that are also property names of the style (instance of CSSStyleDeclaration) JavaScript object.

But that's only a theoretical issue. JavaScript has historically been preventing collisions between HTML/CSS names and JavaScript object property names (the most notable example being class HTML attribute → className in JavaScript). Nevertheless, if ever the user queried such a name, the code would simply make a "false positive" and return the input value as-is, unprefixed.

However, I noticed that when querying a CSS property name that exists unprefixed, it doesn't get cached into the vendorProps map (contrarily to the prefixed names, that get cached).

Therefore, when querying unprefixed names, each time there is one lookup in the vendorProps map then one lookup in the emptyStyle object, whereas we could save the latter (after the cache is populated on first access).

Namely, I'm suggesting this change:

export function finalPropName( name ) {
	var final = vendorProps[ name ];

	if ( final ) {
		return final;
	}
	if ( name in emptyStyle ) {
		// replace:
		//return name;
		// with:
		return vendorProps[ name ] = name;
	}
	return vendorProps[ name ] = vendorPropName( name ) || name;
}

Then to avoid confusion, we should rename the vendorProps map to something like finalProps.

(Just as a reminder: The user might also query already prefixed names. Such use case would, and should, work as well. Just something to keep in mind.)

I opened an issue about this: #5582.