using Search looks like the simplest solution as you're loading from a database, and the MassIndexer will do most of the hard work.
With Infinispan Query there is no MassIndexer, so besides loading from database yourself you'll also have to make sure you can scan the full table at startup. Not hard at all, but it quickly becomes more complex if want to minimize bootup time using multiple threads for the initial indexing: they speed up a lot with Hibernate Search and a similar feature is not available on Infinispan Query yet (contributions welcome and not too hard as Search already provides most reusable logic).
Using Hibernate Search seems the simpler approach; in case you wanted to try Infinispan to take advantage of in-memory caching, just enable a Infinispan 2nd level cache for Hibernate.
thank you very much for your answer.
It turns out the tables are not that easily mapped to the actual searches, and in the name of decoupling, I am currently experimenting with option 2. In neither solution can I get infinspan query working in JBoss 7, but I'll post about this separately.