There is a difference between wave function collapse and destroying a photon. Normally, when you detect a photon, you let it hit a device, where the photon is absorbed and its energy is transferred to an electron that creates a photocurrent you can measure. So, after the measurement, the photon is no more, you only know that you had one but cannot do anything with it now.
The thing that Serge Haroche does is that he measures photons without destroying them. In this way, he knows how many photons there are in the cavity, and can perform other operations on them later. But of course, when he starts with a coherent state and measures the photon number, the state will collapse into a Fock state.
The basic idea of this measurement is to use dispersive interaction with single atoms. Thus, the photons are not absorbed, but imprint a phase shift on the atom that can be measured provided the atom is initially in superposition of ground and excited state. If you want to know more about the technique, I suggest you look at articles describing these experiments (from Nature and Phys. Rev. A).
Edit: arXiv versions of the mentioned papers: http://arxiv.org/abs/0707.3880, http://arxiv.org/abs/0905.0114.