Oh, vocal processors can absolutely be profitable, but only for companies who have enough engineering resources to focus on them. If we made a vocal processor, that's another non-vocal processor we can't do, and our metrics consistently tell us that non-vocal processors are more profitable for us.
Variax is another example. Yes, we'd love to develop a next-gen Variax, but we'd like to pay our mortgages/rent and feed our families even more.
Roland/BOSS used to show over fifty new products at NAMM, many of which were very sophisticated. They have the ability to experiment with digital accordions, digital Chinese fiddles, digital Clarinets, etc. and even occasionally release test market products they likely know won't sell. Yamaha Japan has even bigger resources. For better or worse, YGG/Line 6/Ampeg/Cordoba needs to focus on very specific markets. (Or markets that aren't being served well—or served at all—by anyone else.)
BOSS vocal processors don't generate harmonies based on realtime guitar playback, do they? I haven't looked in a while.