Sun's fate was always a concern before Java was GPL'd, but it is an entity in its own right now. So, whilst Sun is the lead cheerleader, should they fail there are enough big companies (IBM for one) heavily invested in Java that one of them would take over the reigns.
Sun should focus on producing Java acceleration hardware, especially making multicore behemoth servers designed for Java thread concurrency. Sun already has enterprise customers, they own Java, and Java is their customers platform of choice.
Sun already has multicore behemoths - it's the CoolThreads platform (UltraSparc T1 and T2 processors). Dozens of cores - and I've used them and they run concurrent Java really, really well.
As for specialised hardware. Sun used to produce/specify the picoJava, microJava and ultraJava processor architectures (processors designed for Java).... They all tanked. The problem wasn't the implementation - it was timing. They came about just when the processor world was coming to the conclusion that the instruction set architecture didn't matter - AMD proved this by bolting x86 onto a RISC core (proved it by doing this, and then by winning the performance war for a number of years with it).
Problem now for Sun is timing again - the market combined with their market share. It will be a challenge to hold onto the market they've got, and a bigger challenge to grow into other areas.
I wonder if "it's open source, this will go on irrespective of corporate situation" is a powerful retention/new acquisition argument for Sun (for Java and OpenSolaris) -- and could ultimately what keeps them alive amidst fears.
