This was to work around CCBC-1204, which Sergey believes has the same root cause. Once that fix is in, lets revert the workaround and use the cluster instead of an open bucket. Be sure to make a test that verifies it.
This is in 3.0.0, but the fix may not happen in that time frame. If it is, this should be simple to undo the workaround. For all of them, we need to insure the tests verify that no open buckets is ok. Whether the tests bring up a new cluster (which we can not open buckets prior to testing), or changing the framework (probably not worth it), this needs verification tests for sure.