From a Kuhnian point of view, the reliability of theory depends upon the system by which it is negotiated.
Kuhn proposes that theories are based upon paradigms, which are simply slightly more general theories. But paradigms are agreed upon by communities, and get re-evaluated on a regular basis when a truly new idea cannot fit with the existing course of the theories currently in use. So the negotiation process by which the community dismantles old paradigms and vets new ones is the basis of the reliability of the theory.
Science as system, then, is the producer of theories. But science itself is not a theory, it is a political process. Paradigms are judged by a broader philosophical construct that relies upon human intuition and negotiation to evaluate what is more likely to be productive and what is less likely.
Of course, that political process is going to depend upon human experience, and that is related to practice.
But the basis for negotiating paradigms is not the practice itself, only the experience of practicing. Otherwise we would never abandon an old paradigm no matter how hard it got to use it. Practice would always encourage the retention of the theory upon which it is based, and ad-hoc rejection of ideas that don't fit. But we have witnessed historically how paradigms shift.
For instance, Galileo's argument that physics needs to have a single set of rules for near and far objects, for mundane and celestial observations, was a paradigm shift. It took away the foundation upon which astronomers to that time based their calculations, presuming the circularity of celestial orbits, but using different conic sections for flight paths near the ground.
A critical mass of people needed to be convinced that this unified system was intuitively more likely to net better results in the long run, (without seeing what it was going to look like when it was finished) or folks would never have given up the precision of the existing system of epicycles that had been worked out over centuries before then (and would never have started working on theories based on Galileo's new assumption).
We have some historical record of how this actual process happens, and a revival of the History of Science in the 1960's led to interesting theories about how things work when a science is reformulating its own basic theories. A very detailed and interesting account of Galileo's success is laid out in Feyerabend's "Against Method" (with strong emphasis on its political nature, and its lack of objectivity). And shorter accounts of many other paradigm renegotiations are recounted in Kuhn's "Structure of Scientific Revolutions" (with less political spin).
