Observability: the three pillars aren't enough — what's the fourth?

the 'three pillars of observability' logs, metrics, and traces are foundational, but i increasingly feel like they're not enough, especially when debugging complex performance issues in distributed systems. they tell you *what* happened and *where* it happened, but not always *why* it happened at the code level. that's where continuous profiling comes in. tools like pyroscope or parca, which continuously sample cpu and memory usage across your services, can show you exactly which function call is suddenly taking 3x longer or allocating too much memory. correlating this with traces and metrics feels like the 'fourth pillar' that truly completes the observability picture. i'm curious if others are starting to see continuous profiling as essential. how are you integrating it into your existing observability stacks? what challenges have you faced in adopting it, and what have been the biggest wins? i think it's the missing piece for a lot of performance investigations.