Workstations with x86s pretty much eroded their low-end Solaris/SPARC market because even a low-end SPARC cost twice as much as a pretty decent x86 running Windows NT. That was the largest mistake - to allow NT to take a foothold in the low-end workstation market. This caused lots of sophisticated software being developed for Windows instead of Unix.
It's probably too late to develop a minimalistic low cost, sexy, sleek box running OpenSolaris out of a T1 or T2 chip, but I would consider such an all-or-nothing strategy to put lots of very cheap, conceptually advanced computers in the hands of students and researchers so to foster development of software that runs better on Sun architectures.
It's a good kind of vendor lock-in: one achieved by excellence. Sun is uniquely positioned to do that.
I hope they do. An x86-only world is dull.