[rigel-mintaka]

Summary

This PR adds the design record docs/designs/observability/compass-go-trace-hops/design.md. It answers Matt's ruling on RIG-5276: "need traceparents everywhere, make sure we're passing it through all services/RPCs".

The trace-logs record (#2219) puts a span on every inbound request. This record covers the outbound hops that still drop the trace:

• H1: guest control client spans (runner → guestd), with options passed in from main.
• H2: otelx.HTTPTransport on every server HTTP egress (forge, Linear, Trunk, linearagent, auth verifier, object store). url.full query values are redacted.
• H3: a pgx tracer for Postgres spans.
• H4: the fabric DLQ keeps the traceparent.
• H5: the in-process bus takes ctx on every publish.
• H6: bus tails and presence edges restore the trace.
• H7: a gate, so a new uncarried egress fails CI.

No proto change. Tasks stack on the trace-logs stack (T9, #2270).

Do not merge yet. Open Question 1 (third-party egress: inject the header or record a span only) is load-bearing and waits on Matt's ruling in RIG-5398. Open Question 10 (a traceparent on runner frames, one internal proto field) is in the same issue.

Verification

• rumdl check is clean.
• Binary sizes are measured with CGO_ENABLED=0 GOOS=linux go build -trimpath -ldflags='-s -w'. The record's Placement section uses the feat(runner): span AgentGateway RPCs and strip stream spans (RIG-5313) #2259 split of provider setup into internal/otel/otelsetup: the CLI is 13,709,472 bytes at the stack tip, and otelhttp in otelx adds 8,192 bytes.

Risks

• No code changes in this PR.
• The implementation adds client spans and Postgres spans. Under the RIG-5373 spend freeze it does not merge until hard spend caps land.

Compatibility

• No public or internal wire change.

Documentation

The record is the spec.

Spec-impact: none

Ledger-impact: none — design text only; no decision-ledger row changes.

Refs RIG-5321, RIG-5186

[trunk-io]

Merging to main in this repository is managed by Trunk.

• To merge this pull request, check the box to the left or comment /trunk merge below.

After your PR is submitted to the merge queue, this comment will be automatically updated with its status. If the PR fails, failure details will also be posted here

[linear-code]

RIG-5186

RIG-5321

[github-actions]

Compass engineering docs preview: https://compass-ux-5321-trace-hops-d.compass-eng-docs.pages.dev

Deployed from compass-ux/5321-trace-hops-design at cf476cb.

Changed pages:

• docs/designs/observability/compass-go-trace-hops/design.md

[rigel-mintaka]

Review result

• reviewedSha: cf476cb70117ad90f408b3b8494fb94a0e4ed8b0
• reviewedBase: 52c78ef0bc63bc62baa83eea991a4ed7d4290748
• Result: all-clear (0 high, 0 medium open)

Notes: 0018da4 records Matt's RIG-5398 rulings (OQ1=(b), OQ10=(b)) and adds H8. Review found one medium: H8 had no path for the Reload trace on the shared publisher stream. cf476cb passes the traceparent per frame through sendLocked/forwardState and tests both shared and one-shot streams; re-review 0 findings. Field 5 is free in PublishEventsRequest.

Gates: rumdl and design-ledger-gate pass; pre-push hooks pass.

Promote to ready once CI is green; merging it freezes the amendment.

[rigel-mintaka]

Moved to RigelBuild/docs#18. Design records now live in the docs repo. The record is unchanged except that its two parent links are absolute URLs to compass main.