JWST was designed for future refueling capabilities (there is a docking target and an externally accessible port for access to propellant storage), and this mission is well within existing launch vehicle capabilities (if a robotic refueler is lifted as payload). At launch, JWST was fueled with ~301 kg of propellant, so subtract that from launch vehicle payload to get an idea how heavy your refueling vehicle can be (assuming roughly 4k-5k kg depending on launch vehicle), and some margin for the refueler to separate and propel itself over the gravity edge JWST balances itself on from L2 into deep space post refueling.

https://space.stackexchange.com/a/38415

https://time.com/6127003/webb-space-telescope-discoveries/

> Webb faces other operational challenges, however. Hubble has been kept alive in part through maintenance and servicing runs done by astronauts. Webb’s great distance from Earth makes that kind of house call impossible. What’s more, in order to remain stable at its gravity-balanced Lagrange point, Webb needs a thruster system, and a thruster system requires fuel. The telescope will launch with a full tank, but that will be only enough to keep it operating for a minimum of five years and a maximum of 10. In theory, a refill ought to be possible, and the telescope is actually equipped with a docking target to accommodate an incoming spacecraft that could conduct a refueling and extend the JWST’s life. It’s an appealing idea, especially considering the telescope’s dizzying price tag. That spacecraft, however, does not yet exist, though it could within the next decade.

To be honest, sounds like a cool project to start working on, with the idea that it becomes an off the shelf capability for future space observatories. SpaceX's Dragon can carry a 6000 kg payload, and already supports autonomous docking with the ISS, for example.

By "servicing" I actually meant more "repair". The initial burns left enough fuel that JWST is likely to wear out before it runs out of fuel. The refuelling I meant was for craft to get people there and back, because repair is significantly more difficult than refuelling.

https://blogs.nasa.gov/webb/2021/12/29/nasa-says-webbs-exces...

My mistake. In that case, it'd likely be more reasonable to get a vehicle to the observatory, boost it back to Earth orbit, and then attempt to return it to L2 after on orbit servicing. I don't know what that Delta-v budget looks like, although I'm happy to throw coffee money someone's way if they want to do the math in this thread.