It's hard to say exactly what to do. If you look at recent history (30 years) you might easily conclude that Squeak is a lost cause. There are too many examples of well-designed, great software apps that have languished in US schools.
When I think about this kind of situation, I think of what I call the "Eazel Effect." Basically, Eazel was a much-ballyhooed company that took a bunch of venture capital to write Nautilus, the file manager for the GNOME desktop, released 1.0 and promptly folded. As a business, Eazel was an unambiguous failure, if they were writing proprietary software, their software would have, at best, been purchased and folded into some other product and at worst simply abandoned. Because they were writing free software, Nautilus continues to be a key element of desktop Linux. The company's failure had virtually no impact on the software's continued distribution, development and ongoing utility. When I got into free software, the Mozilla Project was widely regarded as a bloated, behind-schedule failure. Once they released Firefox a couple years later, all was forgotten and forgiven.
Similarly, Squeak would have been killed outright by both Apple and Disney if it was proprietary software, and it would have subsequently died of neglect as a commercial product. But free software doesn't run on the same schedule that commercial software does. As I've mentioned, a host of the most exciting new projects in ed-tech use Squeak: OLPC, Scratch, Croquet and Kusasa. And it is really a small leap to imagine a world where the next time the MacArthur Foundation pours $50 million into digital media and learning they structure their investments to leverage these existing free cross-platform resources and initiatives, instead of subsidizing both the creation and sale of proprietary software (which seems to be their model now). The thing is that Squeak is both a small scale "tinker-toy," allowing students to extend it and create little toys and applications, and a large-scale "tinker-toy," allowing software engineers to build complex components atop it.
Now, I'm not predicting Squeak world domination soon. Safer money is on a similar set of tools developing on Python, a whole different set of tinker-toys (and the one I actually use).
It is a truism that, while writing software is hard, getting your innovation into schools and used effectively is even harder. With proprietary software, when we get the marketing, distribution, training, etc., wrong, everything goes away. We need the fault-tolerance of free software, where multiple vendors and distributors get an equal longer-lived crack at getting innovative software in the hands of kids of teachers.