I think the idea of different NVIS antenna designs for different situations is pretty solid, so I have been thinking about parameter ranges for various situations.
Permanent install:
It’s worth noting that in a pinch a ham constrained by HOA restrictions could use a portable NVIS antenna solution and only erect it on a temporary basis, if putting it up is sufficiently easy and fast.
One really appealing solution would be a horizontal loop, full wavelength on 80m or even 160m, up about 10m AGL and fed through a remote tuner. Plenty of band flexibility, and great NVIS performance, plus a bonus of being usable on 20m and higher.
But the obvious easy solution for a permanent installation is an OCFD cut to cover 80m and 40m (and presumably 20m, 17m, 15m, 12m, 10m) Such an OCFD could be constructed pretty cheaply, and there are numerous commercial versions available, including my favorite, the nearly bombproof Buckmaster OCFD. Hang one of these as an inverted V with the apex up about 15m and you are set. The only complication is the need for a 4:1 current balun.
In my mind, the big advantage a permanent installation can have is it’s possible to do something to improve the ground quality. You can improve the ground invisibly, so that could be useful even if you need to not have the antenna permanently installed, but instead erect/strike it as the need demands.
Portable install:
There’s a certain amount of convergence there but not so much that I think it’s not worth exploring different designs.
The simplest possible NVIS design is a simple dipole at about 4m AGL. Sadly, this rules out a link dipole, where you open/close links to change bands, because the links would be up out of reach. If you’re willing to have the thing low enough then you can open/close the links. Perhaps a solution would be a trap dipole. (Suggested bumper sticker: “Free the trap and cage dipoles”)
One thing I haven’t really explored which I think ought to be looked at closely is a loaded dipole, perhaps a fan. How short can you make a loaded dipole before it is just a dummy load?
One final problem which rears its head for both permanent and portable installations differently is bandwidth on the 80m band. 80m runs from 3.5MHz all the way up to 4.0MHz, and it’s darn hard to get a decent SWR across the entirety of the band. Sure, we’ve got a lot of tools in our bag of tricks - cage dipoles and variants like butterflies, and even sneakier tricks like sleeve dipoles where you put another parasitic element close to the dipole that will be capacitive for part of the range and inductive over another part. But all of those tricks basically increase the sensitivity of the design to small construction errors, so actually implementing those designs becomes challenging.
Maybe ‘difficulty of implementation’ ought to be an entry in the design goals.