E04-TS: Yes
Rationale
External reporter Michael Renner (terrorobe) published quantitative Tailnet Lock authority-reconstruction measurements for Tailscale 1.98.9, with an explicit synthetic 825-AUM generation recipe, filesystem configurations, timing boundary and allocation counts. Linux amd64/ext3 measured tka.Open at 5.871 seconds and 1.98 GB cumulative allocation; tmpfs measured 5.468 seconds. PR 20736 publishes the fix and benchmark machinery. This is a reproducible measurement of a selected client mechanism, satisfying E04 without requiring every performance dimension.
Assessment method
Broad primary-source and pinned-code review; no new runtime test
Source context
Round 597 primary-source review
Outstanding qualification
Independent execution is attributed to the external reporter, subsequently a contributor of the fix, not independently rerun here. The measured 825-AUM workload differs from the merged 2000-AUM regression benchmark. B/op is cumulative allocation, not retained memory. Historical results do not describe current patched performance, packet throughput, managed-coordinator capacity, global scale or the macOS behavior of the selected Linux profile.
Public sources
- https://github.com/tailscale/tailscale/issues/20735
- https://github.com/tailscale/tailscale/pull/20736
Editorial evidence references
The following identifiers refer to research records not included in this public package; they are not public links.
- round597.html
- round597-sources.json
- round615.html
- round615-sources.json