Speaking out of my rear but with some confidence that this guesstimation is more or less right, here it goes:
The total cost C (in terms of time only) of switching to Java is proportional to the avg. cost avgC (time) of getting over the language/toolchain learning curve multiplied by the maximum number of edges in a network of n programmer (the burden of communication and collaboration), not amortized by each individual's familiarity with the new language (otherwise, we'll have to play with weighted graphs and get rid of avgC)
C = avgC * n(n-1)/2
If it is only you, then n=1 :)
If terms of risk, you take that and multiply by a risk factor R specific to the size and business importance of the application.
C = avgC * n(n-1)/2 * R
If it is a pet/exploratory project to learn the skills, we can say that R approximates 1 (but it is not one because there is still a risk of failing or doing poorly and thus not getting the desire skills.) If it is a real, production system, then depending on the size, R increases (possibly exponentially).
Putting all that for-fun-but-not-really puzzle aside, the cost of trying Java (or any other language) on an exploratory project with minimal risk should be minimal to you. And it should only involve learning the language and the toolchain (which are very similar to what you know already.)
Go for it ;)