Guess the community (or a company like RedHat) will have to step in and fork 2.7. There is no way that people who haven't switched, will want to switch, unless absolutely forced to (for example: no security patches).
The vision behind crypto is more decentralized than the actual currency is. Some may applaud the SEC stepping in to combat shady ICO's, others see it as overreaching death throes by an increasingly anxious dinosaur.
Bitcoin was set up to disproportionally reward the early adopters. The early adopters were cyberpunks, cryptography enthusiasts, and darknet merchants. Since we can only speculate about the inventor(s), the motivation can range from a US-hostile government wanting to loosen the grip the US has on fiat and digital currency, to Libertarian Anarchists wanting to switch trust from bank-bailing-out governments to technology.
It seems one goal "to protect Bitcoin from the select few" is already close to impossible, as Bitcoin shows a highly skewed power distribution (a select few has thousands of Bitcoin).
Try looking for a company that does take-home assignments (these are mostly Kaggle-type problems). Then you don't need to learn formulas by rote, and the interview can focus on the assignment you did.
Experimental design (statistics) is important, but data science is a team sport. The people specialized in statistics may not be able to code up a deep net, and vice versa.
Computer scientists make great data scientists, especially if you know data infra. If I were to start a team from scratch, I'd focus on software engineers and physicists. Most are willing to pick up some ML to broaden their skills, and the ability to go end-to-end from conception to production is invaluable.
> 1) Has anybody run into a request like this? How would you respond?
Yes. Satisfy their justification without showing code. It is possible to show how algorithms made their decisions. Make this part of your product. You already noticed there is a demand for it and that delivering a black box can be a deal-breaker. So read up on LIME, decision paths, interpretable models on black box output, etc. and give them the capability to see how an algo made its decision.
> 2) How likely do you think their goal is to genuinely "see what happens under the hood" as opposed to replicate in the future?
Unlikely they'll replicate. It would set them up for legal problems. Depending on how deep your moat is (training data, novel optimization techniques, encoded domain expertise), they probably wouldn't even need to see source to replicate in-house. It may be more about not being hood-winked, paying top dollar for a product that does a few imports from open source libraries.
> 3) Are there any legal protections we can put in place to prevent them from not just copy-pasting our code, but also from "learning from it" or so?
Not that I know of. Perhaps you could charge extra for the code review, so in the case of "learning from it" they'll at least pay for it. A thought exercise: did you learn from open source/open research/commercial solutions before building your Proof-of-Concept? If no, then they don't need to either (provided they can hire the talent), if yes, you are like a thief who is worried they will steal from you :).