The people they send to meetings on a regular basis do a great deal of important work, and they often disagree among themselves. But sometimes marching orders evidently come down.
There was a recent concerted effort to define a process to determine when and where backward ABI compatibility should be abandoned, which would have made it much easier to bring about such occasions.
Maintaining a list of individually unobjectionable principles seems innocuous, but it is via the principles omitted from the list being thereby made harder to appeal to that you can get bad results. And, the more there are, the easier it is to reject this or that request.
However, with the exception of some communication missteps around the licensing changes (which actually ended up in fully open source JDK, as opposed to the mostly open source JDK in Sun's time), Java has actually done great under Oracle.
Amazing what having resources can do for a project, eh?
I mean, I'm not thrilled with the licensing model around GraalVM (since it's another "mostly" open source situation), but business is business, I guess. I'd personally prefer that they treat it the same as Java proper, and fund it via support contracts/subscriptions.