Thermoplastic fracture is a multiscale process where nanometer-scale polymer chain dynamics drive micrometer-scale mechanics. Current research typically relies on molecular dynamics (MD) to capture molecular mechanisms or finite element (FE) methods to model bulk mechanics. However, MD is traditionally restricted by spatial limits, while FE models typically lack molecular fidelity. While multiscale techniques are rapidly evolving to bridge this gap, their validation is hindered by a scarcity of high-fidelity data that simultaneously captures both nano- and micro-regimes both on the experimental and simulation side. This contribution addresses these challenges by directly connecting molecular architecture to microscale failure. We developed a framework that upscales an established MD model of a generic thermoplastic, enabling simulations of up to 30 million superatoms. This approach captures multiple micrometers of material, significantly exceeding previous MD limits. We systematically varied chain length, entanglement density, angular stiffness, and boundary conditions to observe their influence on crack propagation. Our analysis reveals that crack propagation is not primarily governed by a minimal pre-crack length; instead, it is highly sensitive to the interplay between boundary conditions, the resulting stress state, and the domain size. Furthermore, we identified and characterized an inactive "dead zone" adjacent to the crack tip. This region acts as a structural bottleneck that must be overcome for sustained growth - a key nanoscale mechanism sporadically observed but previously seen as peripheral data. In conclusion, these simulations provide the high-resolution benchmark data essential for validating next-generation multiscale modeling. By linking molecular topology to micrometer-scale failure, this work offers crucial insights into the fundamental mechanisms of thermoplastic fracture.