A substrate that can take a polish like glass, weighs as much as carbon fiber, and rigid enough to hold shape without an elaborate cell.
The mirror and the cell are to the two highest density/highest weight components in telescopes. Find a substrate to replace glass and then eliminate the cell - that would be game-changer.
Hmmm... not sure the material you describe would be as big a breakthrough as you might think. The key parameter here is "specific stiffness", the ratio of modulus to density. Using the units employed by Table 6 from "The design and construction of large optical telescopes", (p. 139, 2003), which are GPa/(kg/m^3), the stiffness of Pyrex is 0.032 and Zerodur is 0.036, while bi-directional carbon fiber/epoxy specimens range from about 0.040 to 0.055 depending on details of manufacture. This is an improvement on the order of 50% (Pyrex to mid-range CF, or Zerodur to top performing CF), which would help but not really be a radical game changer by itself. (The book has the two extremely costly exotics silicon carbide and beryllium holding pride of the show at 0.146 and 0.162).
Now what would be more radical is if it were economical to get away from the constraints of a homogeneous flat mirror - meniscus mirrors, hollow core, foam core, ribbed mirrors, or other designs that create hollow frames of some sort. Could a composite type of mirror substrate help there?
I would say the Hubble Optics has made a stab in that direction with their low-cost sandwich mirrors, though their sandwich technique is fairly crude. There is quite a thick forest of glass columns on the two halves of the mirror blank, which in my mirror don't match up very well - I think it is over-designed using extra material to substitute for finesse, a more careful process should be able to get much better weight savings.
And there is that StarStone project out there working on foamed mirrors. They say they are shipping 18" and 24" blanks and "light bucket" (worse than 1/4 wave) mirrors - has anyone seen one of these?
And on point 2 - the cell. What is a typical cell weight to mirror weight ratio? You can of course make a cell arbitrarily heavy if you like, but it seems to me that the cell need not be more than a fraction of the weight of the mirror since it is free to use its materials in a more efficiently stiff structure than the flat mirror (i.e. it should already have the structure we would like in an advanced mirror blank). An advanced mirror might do away with the need of a sophisticated cell, but the cell shouldn't be a big weight driver compared to the mirror itself.
What is the flexure tolerance in the cell anyway? It provides floating support points, which are designed to inherently adjust to distribute the load. The main problem would be collimation shift due to rotation of the support plane (and the mirror) relative to the secondary as the altitude changes, and only uneven changes in cell flex would cause this. Am I thinking about this correctly?