Upgrading to Logstash 2.2 involves significant performance improvements and changes to the threading model, which now unifies filter and output worker threads into a single 'pipeline' worker thread type. This new architecture can lead to increased use of file handles, and requires tuning of the -w setting to optimize performance, potentially involving more threads than before but with improved efficiency due to less context switching. The introduction of a new batch size setting -b affects the default flush size for the Elasticsearch output, with the flush_size option now only setting the maximum flush size. However, an issue with Elasticsearch output workers not sharing the same backend connection pool can lead to an excessive number of connections, necessitating a temporary workaround by reducing the number of Elasticsearch output workers until a patch is released. Users are encouraged to provide feedback on these changes through various channels including Twitter, forums, and GitHub.