I would say SunOS (i.e. pre Solaris SysV and not including it) was the quality bearer for UNIX in that era. Particularly once they did the Unified Buffer Cache; HP-UX was never able to accomplish this and it makes it not an ideal file server amongst other problems.
HP-UX 10 and 11 progressively imported more SysV code and lost some of the charm that 9.x has.
I find AIX to be fascinating. Especially 3.x against contemporaries with its LVM, and a pageable kernel. A lot of people have snap judgements against it because they saw 'smit' but don't really understand anything about it.
SunOS evolved to be a great file server and "network computing" server, whereas HP-UX evolved to be a better multi-workload commercial server. Horses for courses—and they often ran on different turf.
You're also right to shout out out some of the other innovators: Data General's DG/UX did a great "let's redesign the kernel for multiprocessing and NUMA." IBM's AIX had kernel threads, pageability, and preemptibility at a time when no one else did (plus JFS, LVM, and eventually LPAR isolation). And Sequent DYNIX/ptx had some impressive multiprocessing (RCU) and large DBMS optimizations very early on. HP was by no means alone trying to engineer away Unix' early weaknesses.
HP-UX 10 and 11 progressively imported more SysV code and lost some of the charm that 9.x has.
I find AIX to be fascinating. Especially 3.x against contemporaries with its LVM, and a pageable kernel. A lot of people have snap judgements against it because they saw 'smit' but don't really understand anything about it.