The scenario I anticipate is the one where Oracle adds something generally useful to the JVM as a commercial feature. Off the top of my head, there's stuff in the memory management area such as a better GC or access to off-heap memory, or goodies for dynamic languages such as tail call optimization.

Suddenly, putting a bit of money in the JVM allows you to go faster. In the beginning, the community will be disciplined about it and make sure that libraries and applications that use these features degrades gracefully, so you can make use of them without getting the commercial JVM. But already here it starts smelling, there are now two target JVMs. Testing just doubled in complexity.

Soon, small, clever libraries that does immensely useful stuff with the new features appear - like what Google Collections did for generics - and pretty soon you're wilfully handicapping your developers by not using commercial JVM. When you already have a revenue-generating code base, throwing Oracle a little money is a no-brainer. There's no way they'd make licensing so expensive that it wouldn't get a CTO fired NOT to buy it. But now they've introduced the drag at the start-up level which I'd voice concern over in my previous post. This is especially sad at a time when the JVM and even Java seems to be enjoying a bit of a renaissance.