If the customer is throwing around a couple million dollars of NRE, it's probably worth bolting something on and pocketing the money. If the customer is only moving a couple hundred thousand, it's probably not.
This seems more like a software-only startup issue. We have hardware and have these hard-nosed discussions all the time--generally about volume.
Customer--"We want X and we want 100,000 units and we're going to put the NRE on this payment schedule."
Us--"No, you're going to put all that NRE up front and it's <2-5 million dollars depending upon change complexity>. And you're going to front the first 10,000 units which is <generally a nice chunk of a million>. After that, we can put you on the royalty schedule. We can't put ourselves in the position of losing the company because your volume never materialized."
Customer--"<choke> <splutter> That's preposterous."
Us--"No. It's the reality of hardware. We know our costs really well."
Side comment: it's always 100,000 units and 100,000 units is generally hardware death--big enough that you have all the volume problems but small enough that your profit sucks.