Investigation stack
Push up to six frames as the incident narrows, pin important evidence, and navigate backward or forward.
Move metric → trace → profile or log → host and back while the investigation stack preserves each frame, time anchor, and entity filter.

The operator should not have to remember every filter they used to reach the current answer.
Push up to six frames as the incident narrows, pin important evidence, and navigate backward or forward.
Use correlation routes and trace/span profile links to prefill filters instead of stitching context only in the browser.
Zoom once and propagate the relevant window into every signal panel opened from the frame.
Use the command palette and investigation bindings to change scope without losing the analytical thread.
workflow
Each step changes the question while preserving the evidence that motivated it.
Capture the service, tenant, labels, and time window that define the change.
Find the failing dependency or span without copying the time range into another backend.
Carry the trace ID and time anchor into structured event search.
Compare a neighboring trace or widen the time window from the previous investigation frame.
proof / shipped
Logs, metrics, and traces share the same storage, time, and tenant model. Profile metadata uses that context while raw pprof remains portable.
Move metric → trace → profile or log → host and back while the investigation stack preserves each frame, time anchor, and entity filter.