Our system is very complex and relies on communication between different services. This is both on software level as on network level.
It may happen (due to different reasons) that a certain signal is not or too late received, but this is difficult to predict and detect. One could design an anomaly detector that observes data and indicates where a problem arose.
Different solutions are proposed in literature: Network anomaly detection methods (http://www.nr2.ufpr.br/~jefferson/pdf/Network_Anomaly_Detection-Methods…), Adversarially Learned Anomaly Detection (https://arxiv.org/pdf/1812.02288.pdf) and Azure multivariate anomaly detection (https://learn.microsoft.com/en-us/azure/ai-services/anomaly-detector/co…).
A literature search can be performed after which such a method can be implemented. Note that one of the largest challenges will be to reliably observe large amounts of data