when does observability become a major bottleneck

hey folks, ive been thinking a lot about observability lately, logging, metrics, and tracing. we're all trying to build systems that are easier to debug and monitor, but sometimes it feels like the tools and the sheer volume of data can actually slow us down. when have you seen observability itself become a performance bottleneck, and how did you address it? are there specific architectural patterns that make observability more of a burden than a help?