Potential Issues with primus.join with Primus-Cluster #74
Description
When trying to add rooms to multiple sparks using primus.join, the process can fail if one of the sparks no longer exists, even though its ID is still listed in a room when using primus-cluster. This issue occurs because the async/parallel method stops if any callback results in an error. This can lead to inconsistent outcomes, where some rooms are joined successfully, but others are not. Is this inconsistency intentional?