I think you could set up an experiment where only one bit escapes from the box, a 1 signifying "alive" and a 0 signifying "maybe dead". But then on observing a 0 the cat would still be in a superposition of states.
Caveat: I am not a physicist. I got 3 A's in quantum physics, but it wasn't at Caltech.
Actually, this would be like the case where you shine a light in the two-slit experiment to see which slit the electron passes through, but the light is dim enough that it won't interact with the electron every time. You get the sum of an interference pattern and a classical double Gaussian in that case.
In the experiment you suggest, the experimenter's wavefunction would be the sum of one where the cat is dead and the experimenter observes a 1, and one in which the experimenter observes a 0 and the cat remains in dead-alive superposition from the observer's perspective. (That is, the weighted sum of a factorizable wavefunction with a non-factorizable one.)
Suppose you believe in the Copenhagen interpretation of quantum mechanics. Schroedinger puts his cat in a box, with a device that has a 50% chance of releasing a deathly poisonous gas. He will then open the box, and observe a live or dead cat, collapsing that waveform.
But Schroedinger's cat is lazy, and spends most of its time sleeping. Schroedinger is a pessimist, or else an optimist who hates cats; and so he mistakes a sleeping cat for a dead cat with probability P(M) > 0, but never mistakes a dead cat for a living cat.
So if the cat is dead with probability P(D) >= .5, Schroedinger observes a dead cat with probability P(D) + P(M)(1-P(D)).
If observing a dead cat causes the waveform to collapse such that the cat is dead, then P(D) = P(D) + P(M)(1-P(D)). This is possible only if P(D) = 1.
If you don't say that only conscious agents can collapse waveforms, then you have to agree that something in the box collapses the waveform as seen from inside the box, while it's still uncollapsed to Schroedinger. And Schroedinger's opening the box collapses that waveform for him; but it is still uncollapsed for someone outside the room. This seems like it might be equivalent to many worlds - all possibilities already exist; you just haven't chosen which one you're going to access until you open the box.
But if you do say that only conscious agents can collapse waveforms, then it's something about their mental processes that does the collapsing. This could mean their beliefs matter. And then, the cat is always dead.
ADDED: People. Read the entire post before responding. I am not claiming that the cat is always dead. I am not claiming that consciousness collapses waveforms. I am claiming that there are only 2 known alternatives:
If you can't produce another alternative, and you don't believe in many-worlds, you owe me an upvote.
Finally, this post is supposed to be fun! You are crushing all whimsy and playfulness on LessWrong when you pile downvotes like bricks on anything playful because it does not provide a complete and satisfactory resolution.