Bifrost 26Q3 Engineering Initiatives
Bucket for Bifrost's engineering-owned work in 26Q3 — non-product-feature work such as
reliability/session fixes, infra/EOL upgrades, and similar independently-pulled items, classed
and pulled independently rather than ranked on the roadmap. Each item is a self-contained story
with its own RFC under rfcs/ — engineering work is documented as an RFC, not a PRD
(PRD = product).
Currently empty — scaffolded for future work. Its first candidate item, Seamless Logout
(cross-product session-invalidation fanout, Jira Story
BIF-8612), was briefly seeded here 2026-07-14 and
then moved to bifrost/adhoc-q3-2026 as story A11 the same day —
it was judged closer to that bucket's existing "small, independent, production-support-adjacent
adhoc" shape than to a heterogeneous engineering-EOL bucket, and Bifrost didn't yet have any other
engineering item to seed this container with on its own. This scaffold stays in place for the next
genuinely non-product engineering request (EOL/dependency upgrade, infra migration, perf/reliability
work not driven by a single product Story).
Scope Changes
- Backend — service/dependency upgrades, EOL runtime migration, infra hardening.
- Frontend — client-side wiring for engineering-owned fixes where applicable.
(Anticipated surfaces — refined once a real story lands; each story's RFC carries the subset it changes.)
Stories — per-story tracker (canonical)
No items yet. When one lands, append the next row as A1 and follow
How the bucket grows below.
| Story | Class | DRI | Status | Added | Scope | BE | FE | QA | Jira Story | Summary (→ RFC) |
|---|---|---|---|---|---|---|---|---|---|---|
| (none yet) |
Progress rollup: 0 of 0 items as of 2026-07-14 — empty scaffold.
Expedite log
Append-only record of the bucket's 🔴 Expedite classifications, once real items exist. WIP
limit: expedite_wip_limit: 3 in flight at once (Bifrost's chosen cap; linter BUCKET_EXPEDITE_WIP).
| Story | Date | Authorized by | Preempts | Reason |
|---|---|---|---|---|
| (none yet) |
Pull rule (deterministic — how an item gets scheduled)
A bucket takes no ## Priority rank — handling is a per-story pull rule off Class / Scope /
owner × capacity.md load × the sprint queue's blocked flags. 🔴 Expedite may interrupt ranked work
(bounded by expedite_wip_limit: 3 + the Expedite log above); 🟡 Standard / ⚪ Intangible never run
ahead of ranked work. Full doctrine:
.claude/reference/adhoc-interrupt-doctrine.md.
How the bucket grows (adding a new item)
- Author the RFC under
rfcs/(type: RFC— context, proposed change, execution plan, risks). - Append a tracker row with the next free
Storyid (A1,A2, …), itsClass,DRI,Status,Addeddate,Scope, discipline owners, Jira Story key, and summary linking the RFC. - Mint the Jira Story under this bucket's Epic (mint the Epic itself first if this is the first real item — one per bucket, never per-AC).
- Mirror staffing into the roadmap
Implementors/QA(union) andcapacity.mdAllocation. - Refresh the rollups — tracker Progress line, roadmap forecast
% donecell, roadmap## Adhoc / Interrupt laneclass buckets.
QA Lane
Lane B — team-policy safe default. Engineering-owned items here (session/security fixes,
infra/EOL upgrades) can carry availability or session-integrity risk reaching paying customers.
Re-confirm per-story with qa-lane-advisor once a real item lands. As of 2026-07-14.