A search data set generated by build "X" of the indexer must be served up by that same build "X" of the search server.

Therefore:

• need to rebuild freedb if the indexer jar changes
• all data sets (freedb + mb) need to be tagged with the ID of the search server war of build "X"
• the action of rotating the data set into place on the search servers also needs to include automatically deploying build "X" of the search server war

The easiest way (but a little bit nasty?) to do that is to drop the .war file into the data directory, then as part of data set rotation, run the deploy-war script (perhaps, if the war file is different from that of the last data set).