Then what is the best solution?
I think it must be using a modular design that allows easy expansion; linearly, instead of exponentially.
Imagine a space telescope made of completely independent hexagonal prism shaped units.
Imagine each hexagonal prism unit also allows easy attachment to another copy from any of its 6 sides.
Imagine the first unit sent to space and so we already have a working telescope.
Then we send another copy and it is attached to the first one (using a drone robot?).
Then we send another copy and another and so on and our telescope keep getting bigger and bigger.
The cost would indeed increase linearly obviously.
(Actually cost of each copy should come down over time.)
Also a big advantage of such a telescope would be much easier and cheaper repairs, compared to a similar size single telescope.
But of course a group of small telescopes attached together would not automatically mean they would be equivalent to a large single telescope.
For that what is called "Aperture synthesis" can be used (https://en.wikipedia.org/wiki/Aperture_synthesis).
Or another solution could be making each unit a "Planar Fourier capture array" (https://en.wikipedia.org/wiki/Planar_Fourier_capture_array).