Well, I don't mean to paint anything as a strong defeater of the project. More of a hurdle, and perhaps not one that everyone would have; the nature of dreams is very individualized and varies so much with so many things. Mine are usually very visual, but if I'm doing a lot of programing for a few days, they become abstract and syntactic in ways I find it hard to recall clearly. Maybe I've trained my dream-recall to overuse spatial aspects as a crutch.

And I would think short and mid-term time contiguity amenable to training (one can certainly learn to spot the skips more readily). And the more common, persistent dream-stabilizing methods would be a good place to start (I find having one dream-hand constantly scratching at its palm, or the palm of the other dream-hand to be a pretty good, persistent I'm-in-a-dream-reminder/stabilizer). Apply one to stabilize the dream, and then keep doing it even with the dream largely stable while thinking through the problem. I didn't try that (when I noticed the time skipping, I started trying to figure out how to use them, like control the skipped interval, rather than reduce their frequency).

My worry is that "filling in what you attend to" is a pretty general thing/feature/mechanism across all the senses I've tested in dreams. Along with "one's expectations", it is near to the foundations of what I think dreams are made of (speaking internally that is; speaking externally, they are made of brain region activations and neuron spikes and such). So there may be a limit to how much one can reduce/corral it on each sense modality. Well exercised short-term memory will probably help with that on the memory channel, since it does on the visual/spatial one... like with objects persisting when you turn your back to them and then look again. But then the capacity limit to working memory comes into play.

And there seems to be only so much I can split my in-dream attention between maintaining awareness that I'm in a dream (so as to keep it stable) and giving room in my working memory for new dream developments/updates to take place. Which is somewhat weird, because if I just look out on a dream vista (or just some wood grain), the landscape can be amazingly complex and rich with shape and color. But the more or longer I control my attention and hold fixed the way the dreamworld is allowed to update (no random stuff insertions or location shifts), the less those richly productive, autonomous visual updates seem to even try. The dream gets duller, and the fewer things around the more unstable it gets. I seem to have to allow the dream production mechanisms some free reign of randomness to maintain the sensation-expectation feedback loop going with content.

But maybe all of this is a product of my outlook. I look for the limits of dreams like the rendering limits of a game-engine. So maybe my dreamworlds oblige me and provide all manner of limits built around the kinds of limits I expect our wet-ware simulated worlds to have. If so, sorry to have infected your dream-thinking with these things.

And heck, just thinking for a bit in a dream about the rough shape, the requirements a solution must have so as to setup some very strong expectations, and then skipping forward in time a bit (or expecting to find the solution in a drawer and opening the drawer) and looking over your subconscious' creative "fill it in as I look" solution could very well be a good tool for seeing possibilities you wouldn't arrive at through more straight-forward thinking.