I have not been able to get this to work in my configuration. The slave instance, starting with an empty target directory, creates segments files for itself on startup. It then has an "index version" that is a higher number than the master's index version. The replication never starts, unless I do a commit on the master.
My replication config is very straightforward. My configuration assumes that the slave machine is starting off completely empty of all data (aside from the essentials necessary to get Solr to start), and will pull everything from the master instance.
NOTE: I tested this with nightly build solr-2009-01-12.zip