That being said, and I haven't done the math, but I think a Falcon Heavy and Dragon could do it at least in terms of delta V; although there are numerous problems with that like the fact that the dragon doesn't have a proper airlock for starters.
It is crazy to think about.
I haven't done the math myself, so I'm certain one side of that equation is wrong, but that seems hard to believe!
[1]: https://mobile.twitter.com/gfish/status/1479339620785524736
In other words, Webb is only one TEN THOUSANDTH the apparent size of a galaxy that's a _billion light years away_
That's just mind numbing to consider!
1) how absolutely mindblowingly big space is
2) how bad humans really are at intuiting things at the scale of space
https://en.wikipedia.org/wiki/Ward_Cunningham#%22Cunningham'...
But I want to reiterate that what's meaningful about this discussion is in part how unintuitive things at the far edges of our scales of perception really are. It's a muscle that, left untrained, will lead you to make incorrect characterizations like the one I made.
JWST is three light seconds from earth
The L2 point is about 1.5 million km from earth [0]. The smallest size Hubble can resolve is about 1/20 arcsecond [1]. It's in a pretty low orbit, so is effectively "on earth" relative to the distance to the L2 point where JWST is. Calculating the size of an angular measure at a given distance [2], you get about 727m [3].
[0] https://en.wikipedia.org/wiki/Lagrange_point#L2_point
[1] https://illuminateduniverse.org/2019/04/11/angular-resolutio...
[2] https://astronomy.stackexchange.com/a/20728
[3] https://www.wolframalpha.com/input/?i=tan%28%281+arcsec%29%2...
Maybe when we have active bases on the moon or Mars would a rendezvous with L2 be feasible but not before. Hubble is in low earth orbit and far more accessible.
On the other hand, with how delicate the optics and sunshield are, a servicing mission might not be worth it. At that point NASA might find it more productive to build a larger telescope, except this time it could again be designed to be manually deployed and serviced, thus not needing the complex automated deployment that delayed the JWST so much.