Fair point, will fix. $900M was just what i could find via google, and looks like it was already rounding up all externalities anyway so will edit. Forgot that to the other posters point around how many cheaper systems will start certifying for human rated/heavy lift deep space payloads, this year will see 3-4 L2 capable heavy launch systems debut with Starship, Vulcan, New Glenn and Ariane 6 (Starship and New Glenn have more risk of being bumped to Q1 2024), so that also makes the case for repair/refit even stronger. Will look into how NASA/ESA manage the L2 decommissioning, you have a point that de-orbiting around an L2 frame much further away from a gravity well is more complicated especially if a satellite runs out of fuel to keep the L2 frame orbit stable. Shouldn't there be a parking orbit established like some now used for very extreme orbit (GEO/Helio)? Likely L2 would want to have 1 type of each highly sensitive observatories (Deep infrared (JWST), Hubble replacement for visible light/UV, High energy, and or LIGO/Radar) for data link throughput and cost reasons given the goals for the new Decadal Science survey that initially proposed JWST, and leave the rest of the workload to the much cheaper ground based telescopes using laser correction. https://www.nationalacademies.org/our-work/planetary-science-and-astrobiology-decadal-survey-2023-2032