I evaluated GraphQL twice before, and discarded it for many of the reasons brought up here. Even this time around, I give a rather lackluster review of it and mention that there are still many caveats. It's not magic, and I'm not entirely in love with it - but it's better than REST.
Query optimization, scalability, authentication, and many other issues raised here were part of the research effort and I would not have moved forward with it if I did not feel that they were adequately addressed.
Before assuming some limitation you have had with it in the past applies to sr.ht, I would recommend reading through the code:
https://git.sr.ht/~sircmpwn/git.sr.ht/tree/master/api
https://git.sr.ht/~sircmpwn/gql.sr.ht
If you're curious for a more detailed run-down of my problems with REST, Python, Flask, and SQLAlchemy, I answered similar comments last night on the Lobsters thread:
https://lobste.rs/s/me5emr/how_why_graphql_will_influence_so...
I would also like to point out that the last time I thought a rewrite was in order, we got wlroots, which is now the most successful project in its class.
Cheers.