wasn't Rock cancelled ?
... you're saying Solaris 11 will rock on the non-rock('ing-bandaid-) T3+ ? In 2013 ?
If you up the per-thread performance of current "UltraSPARC-T*" by 3-5x you're still not close to commodity-off-the-shelf whiteboxes (dare I say x86/x64). Intel's / AMDs x64 CPUs have per-socket cores/threads of about ~20 right now. That's not far off what existing T-series ultrasparc have. So the prospected gain over 3 years is hardly impressive (not compared to existing T3 nor to other CPU suppliers' babies).
It's catchup but not leapfrog. The rest of the world isn't standing still while Oracle/Sun plays catchup.
Solaris as operating system scales to enormous threadcounts , fine, one of the things it's rightfully being loved for. But that doesn't mean your application side has caught up. Scalable software is still a form of (arcane) art, all talks notwithstanding about how Java makes that automatic, how the database does it for you, how the OS enables you to do all that. As you say, it takes years - have your all your app suppliers done their piece yet ?
If your CPU singlethreaded performance is abysmal then one required application exhibiting such a "scaling issue" will pretty much eradicate T3+ from your list of potential hardware to use.
Stuff that scales out has its market, granted, and the T3+ as its existing cousins will shine in those. Whether that can grow to eat into where (previously) Sun use(d) Fujitsu's SPARC64 - that remains to be seen. As it remains to be seen whether it can ever eat into low-end/low-power channel that's currently being stuffed with Atom- or ARM-based bits&pieces.
I hope that Oracle doesn't fall back into Sun's folly and believes its own CPUs will reach competition-trashing performance "when they ship" (especially since Oracle hasn't yet improved Sun's track record on shipping-on-date).
I'm an x86/x64 fanboi - give me Solaris 11 for those please.