It depends...
Theoretically the risk of having multi initiator zone related fabric problems in 3 host / 1 array environment is fairly small, so you might go with a concept of "dont fix it if it ain't broken". Just for an example - I do have 16 hosts here, multi initiator zoning. 14 of them are restarted regularly since this is a Lab environment. We have 2 FC switches and 2 FC arrays. I've never had any fabric issues because of this (however I can't promise that you'll never have).
If you're having some pressure from above or are just loosing sleep because you've not followed the best practices, then go ahead and reconfigure the zones. By this you'll minimize the risk of having IO freeze due to RSCN/LIP events but will increase complexity and management overhead a bit.
hope this helps
WBR
Imants