Code clone detection has been a central topic in software engineering research for more than two decades and underpins a wide range of studies on software maintenance, refactoring, and quality assurance. Despite the hundreds of proposed clone detection tools spanning syntactic, semantic, cross-language, and learning-based techniques, little is known about whether these tools remain executable in modern software environments or whether their reported experimental results can be independently reproduced over time. In this study, we conduct a large-scale empirical investigation of the reproducibility of published code clone detection tools, focusing on tool executability and result-level reproducibility. Using a multi-stage screening process inspired by PRISMA-style study selection and complemented by backward snowballing, we refined an initial corpus of 7,380 clone-related publications to 854 primary studies published between 2001 and 2026. From these studies, we identified and curated a representative subset of 84 distinct clone detection tools for hands-on reproduction, balancing detection paradigms, clone types, programming language support, and publication eras. Reproduction was carried out using a standardized, protocol-driven framework in which independent investigators attempted to obtain, configure, and execute each tool using only the documentation and artifacts provided by the original studies. Reproducibility outcomes were systematically recorded using a unified data collection schema and a graded Tool Executability Status taxonomy. Our results reveal clear and systematic differences across generations of clone detection techniques. Classical text, token, and AST-based tools are generally executable with minimal effort, whereas many semantic, cross-language, and learning-based tools exhibit substantial reproducibility challenges, including undocumented dependencies, environment drift, and result divergence. Notably, successful execution does not necessarily imply reproduction of reported results. These findings provide empirical evidence that reproducibility challenges persist across the clone detection landscape and may intensify as techniques evolve toward more complex and dependency-heavy models, motivating stronger artifact and reproducibility practices.