I took the chance to read back through this thread and had some thoughts: while I agree there are still a ton of rough edges on IPFS (especially on the performance side), I don't think the Radicle case is actually an exemplar of the "ecosystem feelings" overall. Many groups are using IPFS effectively (Haven, 3box, Audius, etc), and other groups like Infura/Berty/Textile get involved in help improve it too (vs treating it as a black box). There are definitely areas we need to keep working on - but that isn't by breaking existing users more, it's by investing in more thorough release testing and network simulation. It's painful right now to slow down to invest in that when we still have a lot of things it feels like we need to fix *right now* - but we've already seen how these frameworks like https://github.com/Netflix/p2plab can help us acutally accellerate development progress, so it's worth the investment. 
postables (Discord) would be happy to chat about your cross-language frustration - I don't think that's something most folks experience so would love to capture that feedback and see what we can do to mitigate.