With silicon technology decreasing in size, memories get more susceptible to external influences, which can lead to soft errors. Although temporary, these errors constitute a challenge for safety-critical systems. Redundancy-based error detection is commonly used in industry to increase safety and mitigate these errors. When an error is detected, safety-critical systems are usually switched to a safe state. While this prevents failures, it negatively affects the system's availability. In this work, we propose Just-in-Time fault detection, a novel method which enables a system to be switched to the safe state only in case a detected error would affect the system's behavior. A software tool enabling the deployment of this method on an off-the-shelf processor is implemented, and the method is validated and compared with a state-of-the-art alternative approach using mixed-critical memories. Our results show an availability gain between 25.2% and 100% compared with the state-of-the-art approach while executing two different standard algorithms.