Metastable Failures in Distributed Systems

Paper review of a phenomenon where distributed systems get stuck in a degraded state that persists even after the trigger is removed. Real examples from AWS, Azure, and Google.

“This failure mode is terrifying because the system looks healthy by most metrics while actually being stuck. I've seen this pattern in production twice and didn't have a name for it.”