At the cost of head-of-line blocking - one dropped TCP packet delays all HTTP/2 streams.
Yes, but head-of-line blocking is a different thing than round trips.
A dropped packet requires a retransmit which is effectively a round trip. Introducing shared fate by bundling many requests into a single TCP connection results in more requests being delayed by round trips per dropped packet.
Isn't head of line blocking a problem with the design of web servers?
HTTP/3 even more so due to QUIC's shorter handshake process.