Simultaneous participant crash impact

Background: Research into participant failures in Federated Learning systems has often focused on the impact of a single participant dropping out.

Question / Future Work: Investigate the collective impact of multiple simultaneous participant crashes within a single round of cross-silo Federated Learning to determine how the resulting data loss aggregates and affects model quality compared to isolated failures.

Why It Matters: Understanding multi-participant failure scenarios is vital for deploying robust fault-tolerance mechanisms that account for correlated failures in distributed systems.

Evidence: Despite that, in a real-world setting, it is possible that multiple participants can crash at the same time due to network failures.