kcs#5170631 explains how to adjust event-ttl via unsupportedConfigOverrides. But reducing event-ttl is a resonable way to help recover from event-spew situations, where the load of listing hundreds of thousands of events causes etcd and Kube API server instability that leads to more events. I dunno if we want to support raising the TTL, but having a convenient, supported, documented knob for dropping the TTL while recovering from event-spew incidents would be helpful.

Alternatively the Kube API server could lean on the fact that Events are informative, best-effort, supplemental data and reject or blackhole new event submissions when it felt like it was dealing with capacity constraints.