A single observation won't overturn an established scientific theory.
Repeated observations and testing that run contrary to the predictions of an established theory, combined with a model that provides superior explanatory power, will result in the formation of a new theory that supersedes the old one. This could be something totally new, or something that is a substantially modified version of the existing theory.
The current best Theory of Evolution is known as the modern synthesis. It is formed from the combination of Darwin/Wallace's Evolution by means of Natural Selection, with Mendel's concepts of inheritance. This has since been modified by a number of other discoveries, including work around the various role of genes and molecular biology.
There has been argument in the last decade or so that the modern synthesis needs to be expanded further (the Extended Evolutionary Synthesis), to incorporate other modes of inheritance (epigenetics, horizontal gene transfer, phenotypic effects, ect, ect) to create a theory more encapsulating of modern understanding.
Mendelian inheritance didn't overturn the ToE. Instead, the existing framework was expanded to incorporate the new evidence and parts of the model (Darwin's concept of pangensis and passing on of acquired characteristics, for instance) was ejected because it didn't provide as good an explanation of observations as the new model did.