Yeah I have seen this from time to time as well. You can run the tests 100 times, and it may fail a few times etc. Not sure what happens, if its pojosr or blueprint etc. Though OSGi and Blueprint has a really complicated startup model where all is async, and kinda waiting for each other. And that is dead hard to test throughly.
You can try setting the surefire plugin to fork the jvm per test. Then its a full fresh JVM on each test. This may fix this, in case there is some "left over" when running all the test in the same JVM.
Unfortunately, I tried forking and that did not seem to help.
I noticed the comment below which is just above the 'Gave up waiting.." log statement in CamelBlueprintHelper.java.
// Note that the tracker is not closed to keep the reference // This is buggy, as the service reference may change i think
Which sounds bad although I guess it is talking more about getting the service reference and then it changing later while it is being used rather than not getting it in the first place.