They support a theory, but they don’t prove it. The theory is just an explanatory model and may be used for making predictions or or retrodiction (think of this as a prediction about the past). As more data is gathered a theory may be found to be inadequate for a scenario or range. When this occurs a new theory may be built to include the new data. Usually the new theory is an adjustment of the old theory and not a complete re-authoring. This has happened within physics with Relativistic theories replacing Newtonian ones, or with the evolution of atomic theories (Dalton mode, Bhor planetary model, Rutherford model, and probably some others). Though there are some occurances of theories that get thrown out all together such as geocentric theory being replaced with heliocentric theory, or germ theory and cell theory replacing the theory of spontaneous generation.
“Wrong” theories can still be useful though. Bohr’s planetary model is still taught in grade school (I guess it’s easier to start off viewing the atom as a miniature solar system until one has built up enough understanding to understand more modern models) and Newtonian physics is still used for the relatively slow moving objects we interact with on the surface of the earth (side note: GPS systems are an example of something we use all the time that must use relativistic physics models).
It’s believed that there may be possible to build theories that include or account for observations that are relegated to different theories (ex: combining electromagnetism theory and relativity into the “Unified theory”) and if this were ever done then it could show that the theories we have on both of those areas to have been inadequate some how.