The escalating computational demands of modern science necessitate the use of Infrastructure-as-a-Service (IaaS) cloud platforms, particularly for large-scale, interdependent Scientific Workflows (SWFs). Critical high-stakes applications, such as the modeling of climate-induced seismicity (e.g., the link between rising sea levels and increased seismic activity in coastal regions), require scheduling solutions that ensure not only efficiency but also unparalleled reliability and deadline adherence. This paper provides a comprehensive, systematic survey and novel taxonomy of SWF scheduling algorithms in IaaS environments. The primary objective is to classify state-of-the-art methods and, critically, assess their inherent suitability and limitations when applied to time-sensitive, computationally-intensive research that must account for dynamic, real-world data, such as the 5% increase in seismic events since 2020. We propose a multi-dimensional taxonomy based on primary optimization objectives (e.g., makespan, cost, energy), algorithmic strategy (e.g., heuristic, meta-heuristic, hybrid), and resource provisioning models, performing a detailed analysis on the strengths and weaknesses of key contributions. Current scheduling paradigms exhibit strong performance on traditional benchmarks but reveal significant deficiencies when confronted with the strict reliability and fault tolerance required by critical environmental models, highlighting a prevalent trade-off between cost-effectiveness and system robustness. Ultimately, we conclude that current predictive models are insufficient—not only due to geophysical complexity but also because the underlying resource scheduling and provisioning mechanisms lack the necessary agility and guaranteed performance to accurately and reliably process the influx of high-velocity, high-volume data, underscoring the imperative for a new generation of reliability-aware, deadline-sensitive scheduling.