Guava is great, thats why its used often. And if you combine projects with each other (hadoop, spark, etc..), they bring in their own guava API level dependency. Normally that is not a problem, if you stick to non @Beta methods, the highest version is used.
I did not find it on the spot, but i remember, there was once a written promise made by the Guava people to remain compatible to a certain API version. And then they broke that promise by changing public signatures or removing methods. Vividly i have the Stopwatch example in my head.
