If we dilute the game, what do we gain?
Alternately, if we ignore the potential of gronked-up senses, what do we lose beside some fans? Unless you're aiming for big money with a mass-market game, it's always better to focus on pleasing your core group of fans. We don't really have a core group, though, so I guess that point can be used to argue against anything.
I know I, for one, would very much enjoy working on gronked-up senses, but I agree that preliminary work should always be using simple renderers etc.
With a lot of gronked-up senses, you can easily get something simple prototyped quickly, I think -- an eyespot could increase the amount of ambient lighting based on how much light it sees, lateral-line tremorsense could work by drawing, say, faint red glows wherever something moving nearby is, for smell everyone can leave behind particles which you render, etc.
Not all are more complicated than sight -- for sight, you need to actually build the meshes for everything, which is a lot of work, enough work that we might even release the first version of multicellular without sight cuz it works better for us, for example. Of course, when a prototype is simple enough, it hardly matters if it's sight or echolocation, for example.
Sight is good to start with, because it's been done before, pretty much by every game. But I'm still inclined towards aiming for realism here.
Of course, you already know my stance, so I'll leave it at this.