When we consider two different representations of a C* algebra, physicists generally bother about whether they are "unitarily equivalent". Now, if the representations are different, the Hilbert spaces on which they act are also
different. So I think the right thing they worry about should be called isometric isomorphism instead of unitarity. (A linear operator which preserves the inner product in the sense that one is careful to confirm to the way the inner product
is defined in the appropriate vector space considered!)