A few years ago the OpenSolaris community was growing by leaps and bounds with just about every significant developer climbing onboard and literally millions of OS downloads going to people who just wanted to play around with the product on hardware they already had.
Today the community is much more muted. Downloads continue, enthusiasm continues to build among technical developers, lots of people are moving Solaris/x86 into production roles running free software, but the market for commercial applications on Solaris for x86 just hasn't materialized.
I think it makes sense to look at this differently from the small and large systems perspectives. On the small systems side, nearly every Solaris/x86 user who's moved beyond simple experimentation seems to be either a developer or an open source user - and my impression is that they're generally pretty happy with what they've got.
A majority of those experimenting with it have not, however, moved beyond that level and most will tell you flat out that they see more demand for Linux and little reason not to sell what's selling.
Personally, I think what happened to OpenSolaris market growth was Ian Murdock - but that's just my opinion, not backed by substantive research and therefore not something to bet on. Nevertheless, what I think happened was first that the role Sun gave him contradicted some of the ideals uniting the OpenSolaris community, and second that his strategy of making OpenSolaris more like Linux got things exactly backwards - the right strategy would have been to continue helping Linux move up to the Solaris level.
Talk to Linux users who've also tried Solaris and the first thing you'll find is that most limited their experimentation to x86 - often because they don't have access to Solaris on SPARC but almost equally because most simply have no idea that the differences go beyond single user response and general usability on personal machines and small servers.
Talk to them about hardware self-diagnosis, about large SMP support, about long term process stability and you don't get a lot of comprehension - what you get is surface level stuff: yes, they like stability, and yes they know what SMP means (but think Intel supports it beyond four cores) but they don't have the gut level appreciation for something that just works you get from having been called to the data center at 3AM to stop an exploding crisis already affecting a couple of thousand workers a continent away.
In a sense it isn't that Solaris on x86 doesn't scale - it's that personal experimental use doesn't scale. It's easy to understand, in other words, why someone whose use of ZFS is limited to typing in a command on a PC and discovering that "Yeah! I can make a pool" wouldn't intuitively understand what that gets you in a production SAP environment.
At the small systems level, therefore, what I think is going on is that people are discovering the obvious: with community enthusiam lagging, for most personal and small business uses there's just not enough difference in technology or personal benefit to make bucking the market worthwhile - or, to put it another way, from a practical perspective they're discovering that if something can be done on x86, it can be done just about as well using Linux as Solaris.
The big systems story is very different. Set aside the people who at some point bought Solaris/SPARC and hosed it by applying Windows or mainframe management ideas where they don't fit, and you're left with a fairly large client base that's quite happy running enterprise class systems on SPARC.
For these guys the primary concerns aren't related to technology or systems stability: their stuff works and is therefore pretty much off the executive radar -so what they're concerned about is Sun itself. To them the unrelenting media campaign against Sun sounds like the distant tread of doom - it used to be absolutely "true" that if Steve Jobs walked across San Francisco bay the media would report it as "Steve Jobs Can't Swim" and, right now, much of the reporting on Sun fits that mold.
So what's their response? Most are simply sticking to what they know, but a few are covering their bets: licensing their applications for Solaris for x86 clusters as a way of giving themselves a minimal hassle bailout route to x86 just in case something bad happens to SPARC.
And really, that's what the small users are doing too: they like it, but the emotional incentives to change aren't there so they experiment just enough to claim familiarity if they need to, and leave it at that.