[jweston]

Generally I'd say that going with Java is optimal if your team feels they have a basic understanding of programming in general. The main pros for me is:
1) it makes it a lot easier to exploit version control since it's text based (you can see what was added/removed on a commit-by-commit basis)
2) WPILib encourages a clean design pattern with its command callback structure.
But that's subjective. Some people will likely disagree.

Honestly our team had many students who struggled with learning LabVIEW. This was especially true for those who had already had some exposure to other C-like languages. LabVIEW's wiring and VIs can become a bit difficult to decipher when the code becomes complex. LabVIEW does make UI design pretty easy, though, and it integrates very well with the Driver Station. Of course, you can use LabVIEW for the driver UI and Java for the robot.

Regarding improving performance, I don't believe moving from Java to LabVIEW is likely to accomplish this unless your Java code is designed in a very inefficient way or your programmers are having enormous difficulty grasping how to use it.

Programming is sometimes (often?) asked to make up the difference for shortcomings from design and/or implementation. Changing languages won't fix that. Good game strategy, well thought out robot design, proper fabrication and assembly, appropriate use of sensors, and a good UX can mean huge strides forward for robot performance. Only the appropriate use of sensors and good UX is within the realm of programming. It is not language dependent at all.