I'd refer you to the comments by Steve and Jez, for starters. Clearly they just want to get in, get paid, and get out, and devil take the hindmost. Do you really think, based on what they've written, that they care about an extra $3 in the BOM?
which is the correct procedure if the volume is not huge.
If you can save $3000 in development (and maintenance) cost on 100 units who cares if the BOM is $3 more.

I have a confession here: it took me a good while to learn this.

I can recall a number of cases in which the coders decided they wanted an extra 500% of RAM and 2500% of processing bandwidth (faster, fancier processor).
the discussion here is which of two reasonable choices to make, not hether to go from reasonable to ridiculous.

we didn't have to buy that multi-k-buck compiler.
that is a ridiculous statement. If you spend "multi-k-bucks" ONCE and every time thereafter the productivity of, say, $50/hour personnal is 20% better, it does not take long to recover the cost.

since they're still shipping hundreds of units per week after two years.
You have a tencency to make your "if you do not do it in assembler, you are an idiot" statements without any "volume qualifier". were I to do a one off, my sole concern would be development cost, would that require "an extra 500% of RAM and 2500% of processing bandwidth" so what. If you do not get professional satisfaction from TOTAL [run] cost, but only from "the most efficient, compact, fast code" you need to get your professional pride meter adjusted.

Erik