Well, this is one reason to believe that we also need a quantum theory that includes gravity (in other words, in which the gravitational effects are themselves quantized). In that case, until you observe the gravitational effects, no actual observation happens. Of course, the gravitational effects probably decohere the state anyway.
Of course, this problem only becomes an issue when the gravitational effects are large enough to be observed. But this is exactly the stage at which quantum gravity is expected to take over.
The gravitational effects of a particle on its environment are not sufficient to change the environment in an observable way. If they were, the wavefunction would indeed change. For example, if you put a gravity-based detector beyond a double slit, if the detector were sensitive enough to feel gravitational attraction from the particle, then there would be no interference pattern - the wavefunction would be collapsed. But since gravity is such a weak force, particles do not exert a significant gravitational force on their environment. There may be an interaction, but the interaction is so minute it doesn't alter the environment, and therefore doesn't create information, in any way. And the wavefunction collapse is not about interaciton, but about the generating of information. After all, you can pass photons through a lens or bounce them off a mirror and not collapse the wavefunction - even though the photons have significantly "interacted" with the glass molecules - indeed, in the case of a mirror, they've been entirely absorbed by the electrons and re-emitted. There's no wavefunction collapse, however, because the state of the glass is unchanged by the presence of the photons.