Developers often apply code refactoring to improve the internal quality attributes of a program, such as coupling and size. Given the structural decay of certain program elements, developers may need to apply multiple refactorings to these elements to achieve quality attribute improvements. We call re-refactoring when developers refactor again a previously refactored element in a program, such as a method or a class. There is limited empirical knowledge on to what extent developers successfully improve internal quality attributes through (re-)refactoring in their actual software projects. This dissertation addresses this limitation by investigating the impact of (re-)refactoring on five well-known internal quality attributes: cohesion, complexity, coupling, inheritance, and size. We also rely on the version history of 23 open source projects, which have 29,303 refactoring operations and 49.55 percent of re-refactoring operations. Our analysis revealed relevant findings. First, developers apply more than 93.45 percent of refactoring and re-refactoring operations to code elements with at least one critical internal quality attribute, as oppositely found in previous work. Second, 65 percent of the operations actually improve the relevant attributes, i.e. those attributes that are actually related to the refactoring type being applied; the remaining 35 percent operations keep the relevant quality attributes unaffected. Third, whenever refactoring operations are applied without additional changes, which we call root-canal refactoring, the internal quality attributes are either frequently improved or at least not worsened. Contrarily, 55 percent of the refactoring operations with additional changes, such as bug fixes, surprisingly improve internal quality attributes, with only 10 percent of the quality decline. This finding is also valid for re-refactoring. Finally, we also summarize our findings as concrete recommendations for both practitioners and researchers.