I think you and I agree about the facts of what was presented. Where we differ is in how we interpret those facts.
The demo was [...] not simply optimising the workflow of some developer pre-compile time decision.
Agreed, and I think this speaks to a misunderstanding between us. When I said "the most revolutionary things they were talking about were workflow improvements", I was getting at the idea that they removed the costly grunt work that has been necessary until to now to implement these technologies, thus putting them within reach of all developers. I was not suggesting the "workflow improvements" weren't backed by supporting improvements to runtime technology, nor was I suggesting they necessarily lived prior to runtime. "Workflow improvements" might be a poor choice of words on my part, since I was really talking about anything that made producing a game simpler, so if the term is muddying the waters, I'm fine calling it something else.
They specifically pointed out that level of detail was being rendered on the screen in realtime, not that it was some optimisation workflow.
Agreed, but that's an incremental improvement over what we've had for decades. LOD techniques have been in use since at least as far back as the PS1. Spyro the Dragon was an early example of LOD in video games, for instance. Even continuous LOD like what they demonstrated here has been well understood and put into use for decades with things like progressive meshes.
On the other hand, the thing that is revolutionary is how any developer can now use that technology with very little work. That's the game changer. Bringing that to market obviously required some significant technology upgrades, which is no small feat, but it's also something that's already been available to anyone...with deep pockets and sufficient technical expertise. The revolution is in making it available to everyone, no strings attached.
The demo also specifically pointed out the ability to render shittonnes more polygons in real time than in the past
I don't mean to sound dismissive, but that's table stakes for a new engine. I'm glad they did it, of course, and I don't want to diminish the work that went into it, but it's an expected evolutionary improvement, not a revolutionary leap forward.
By and large, these sorts of improvements raise the floor for what we can expect, rather than raising the ceiling.
that's a strange comment given the high ceiling the tech demo displays over any other game on the market regardless of studio size.
The whole point of a demo is to show how high the ceiling goes, so I'd certainly hope it pushed the ceiling up a bit! But that doesn't mean that it did so to a significant degree, whereas it did significantly raise the floor. High framerates? High polygon counts? CLOD techniques? Global illumination? You can find any of those technologies being used in AAA titles available today, and by the time UE5 debuts (which is presumably sometime after developers get access to it in early 2021), you may even see games available on the market that surpass what we saw in the demo. But do you know where you don't tend to find those technologies? In games outside the AAA sphere. Bringing those technologies to the 99% of developers that couldn't afford them before raises the floor, and it does so in a major way, hence why I asserted that they raised the floor more than the ceiling.
you're still reliant on developers creating proper light maps for any game that is intended to look somewhat decent.
Perhaps I'm misunderstanding, but it sounds like you're basically saying that implementing raytracing is easy (as with pretty much every other Computer Science student ever, I too built a ray tracer in a single sitting a few decades ago, so I would agree) and they already have to deal with the annoyance of producing lightmaps, so adding raytracing for free doesn't really do much to raise the floor because developers who wanted to do so could already add raytracing. Is that a fair restatement of your point?
If so, I get where you're coming from and it's a fair point, but I'll mention these two things:
A) Lightmaps aren't as onerous to produce as you might think. There are build-time raytracers and other tools that can automate the process of producing most of them. But really, their cost doesn't matter much because...
B) Lightmaps will remain necessary for the next few years on hardware where raytracing isn't possible (i.e. the bulk of the market), so producing them is a necessary part of the cost of doing business. As such, companies both big and small will bite that bullet. In contrast, raytracing is a nice-to-have that isn't widely supported, so any cost above zero is more than most small companies will be willing to pay. Their time would almost certainly be better spent doing other things.