Very few folks don’t run some reverse proxy in front of their application servers (or a whole string of them to load balance), so lack of http/2 on an application server is rarely a dealbreaker for a web framework. I’ll take a solid framework with only h/1.1 over an iffy one with h/2 any day of the week.
The FreeBSD project has been dealing with HTTP2 interoperability hell over the last month. (Curl 7.62 may or may not talk to Phabricator if HTTP2 is enabled; it seems to depend on some local conditions and the phase of the moon whether it works or not. HTTP1.1 just works.) Given that experience with HTTP2, I'm quite happy to stick to HTTP1.1 for now.
HTTP1.1 is expected to just work it has been here for a very long time. It takes time to understand, support and adopt new technology. I'm not sure if the blame should be on HTTP2 in this case.
Regardless of who the blame should land on — I don't have a strongly held opinion about that — the point remains that HTTP1.1 is a super old, well understood, relatively simple protocol that everyone has done correctly for years. In comparison, HTTP2 is the new, more complex thing and it's understandable there are growing pains and incompatibilities. I just don't want to debug those issues in something that worked great for me on HTTP1.1.