I believe the GP is correct: wear and tear on existing components means that fuel won't be the only issue in 20 years. Hubble has been serviced 5 times [1] for example.
Part of the complexity (and thus cost) for JWST was all the issues introduced by trying to fit a 6.5m mirror on existing rockets. All those moving parts and points of failure have design costs.
What you may find in 20 years is that we'll want to launch an even larger mirror and/or use interferometry between multiple telescopes to improve effective resolution.
[1]: https://www.space.com/15892-hubble-space-telescope.html#:~:t....)