Interestingly enough, similar models seemed to be more common back in the dark ages of computing before terms like Open Source and Free Software were really a thing, particularly on minicomputers and mainframes. It made its way to the "microcomputer" world in more limited fashion, then sort of slowly died out. But I think there may still be something to the notion of a license that, while not remotely meeting the definition of "free software," still gives you access to the source.

I'll be honest, though; I think FOSS advocates have a tendency to (perhaps unconsciously) downplay how important "free as in beer" has been for the adoption of a lot of major software projects. An internet in which Apache, MySQL, PHP, and other major free software projects were released with commercial "source included" licenses would probably look very different.

I have my own thoughts about RethinkDB's failure as a commercial entity, but despite having actually been inside, I have pretty limited insider knowledge. Perhaps we should have been trying to push an "enterprise edition" with guaranteed support much earlier, following the "charging for OSS stuff" line you mentioned; as others have said, maybe we just didn't have enough avenues through which to give us money. (It's hard not to notice that MariaDB's .com site is essentially page after page of ways to give them money, for instance.)

I think a combo of paid with source and closed-source extensions can still get lots of success. We're already seeing it with enterprise editions for features that basically only enterprises use. That keeps them from crippling the main, OSS product which costs less. I also envision giving free licenses to people that contribute lots of code, docs, testing, etc. One way is to simply not charge businesses below a certain revenue amount or only to charge a support fee tied to cost of delivering support. As they grow, so does the software house like in the royalty model.

Profit? Sure.

Longevity? Not even close.

Things like the Spice3 circuit simulator and the Magic VLSI editor can continue to exist and run 30+ years later solely because they are open source.

However, is MCP actually running directly on modern hardware or is it being emulated through multiple layers?

Of course, you can use my own argument against me that the ports of Spice 3 and Magic to modern hardware are hardly the same program as 30 years ago.

The same reason for a lot of what's in open source. See PHK's A Generation Lost in the Bazaar for some truly unimpressive stuff.

"However, is MCP actually running directly on modern hardware or is it being emulated through multiple layers?"

x86 is emulated on a custom RISC machine. So does it matter when saying whether x86 is a proprietary ISA that still dominates desktops? MCP is proprietary software whose apps keep running for decades regardless of what hardware they're running on. Like other mainframes, the AS/400, whatever. I hope you're not looking for loopholes around the fact that proprietary software can last decades if it's mission critical, supported, and hard to move off of. It can and does in quite a few cases.

"Of course, you can use my own argument against me that the ports of Spice 3 and Magic to modern hardware are hardly the same program as 30 years ago."

I can't since I don't know enough about them. The mainframe and futureproof markets for software try to maintain backward compatibility with stuff you don't change. So, the part the app runs on is probably the same with who knows what underneath. Sometimes that part is ported, emulated, virtualized, etc. Varies. API and software above stays the same.