This presentation addresses recent work at the TCV tokamak aimed at a better characterization and understanding of the dynamics of the burn-through phase employing both experiments and simulations. In particular, the conditions giving rise to a significant amount of fast-electrons during the startup phase has come under scrutiny [1]. After developing a robust startup scenario with strong fast-electron signals, scans in pre-breakdown gas pressure as well as post-breakdown gas fuelling amplitude and timing have been performed. This study leverages a rich gamut of fast-electron diagnostics available in TCV with increasing energy sensitivity: vertical ECE (10-100keV), X-ray spectrometry in-vessel (10-300keV), and X-ray spectrometry ex-vessel (0.5-20MeV), which give an exclusive insight into the energy evolution of fast-electrons through the startup phase. It is observed that prefill pressure changes can have a significant effect on the early burn-through plasma properties and fast-electron signals but do not lead to significant changes in the flat-top fast-electron population. Gas fueling scans revealed that the plasma density has a direct impact on the onset of fast electron signals and can thus be used as an active control knob to avoid startup fast electrons. However, both the amount and timing of the application of external fuelling are important. These experimental scans are modelled using the 0D burn-through code STREAM [2]. This code solves for energy and particle balance fluid equations in 0D including a basic circuit equation and impurities. STREAM includes a state-of-the-art model for the generation and loss of runaway electrons consistent with the presence of partially ionized impurities. New observables such as Langmuir probe data to quantify the open-closed field line transition, early ion temperature measurements, and fast-camera measurements of the early plasma volume allow to constrain simulation input parameters and better validate the modeling results. Simulation results will be presented on shots with and without fast-electrons. Both agreement and disagreements with experimental scans will be shown to highlight the strengths and weaknesses of the STREAM model. These efforts pave the way towards using burn-through simulations predictively in TCV.