GloriousFlywheel Cleanup Structure 2026-04-17
Snapshot date: 2026-04-17
Purpose
Turn the post-reconciliation cleanup work into one explicit execution structure.
This note is the deeper internal companion to ../cleanup-program.md.
Use it for:
- workstream boundaries
- issue-to-deliverable mapping
- merge order and proof expectations
- deciding what belongs in the current cleanup cycle versus later product work
Current Baseline
Repo state after the slice PR wave:
- docs/status/Bazel-identity cleanup is merged
- backend and operator-path convergence is merged
- dashboard auth/control/audit hardening is merged
- dogfood/shared-runner contract cleanup is merged
- no open PRs remain in the main repo
Live open issue set:
#213runner workspace hygiene#215FlakeHub publication RFC boundary#210downstream adoption and migration kit#211GitHub-first forge boundary#212benchmark methodology#66pulp/package-management RFC#44GPU runner docs
This means the cleanup program is no longer “slice the worktree.” It is “execute the remaining issue lanes without blurring boundaries.”
Workstream Map
Workstream A: Runtime Integrity
Primary owner issues:
#213
Scope:
- host-side workspace hygiene on
honey - repo-owned ARC lane runtime truth
- preserve the heavy-Nix proof on
sting - separation of repo-owned versus personal lanes
Expected deliverables:
- one runner-hygiene fix or operational runbook update
- one heavy-lane validation proof point
- one runtime contract update if live evidence changes the public story
Proof expectations:
- live runtime audit evidence
- successful downstream heavy-lane canary evidence
- no reliance on theory alone for runner-envelope claims
Out of scope:
- broad control-plane redesign
- backend-family redesign
- reopening
honeyversus cloud target debates
Workstream B: Downstream Contract And Product Boundary
Primary owner issues:
#210#211
Scope:
- downstream migration kit
- canary expansion beyond the first tranche
- GitHub-first docs and positioning cleanup
- GitLab and Codeberg compatibility framing
Expected deliverables:
- updated canary inventory and tranche state
- tightened downstream examples and docs
- clearly bounded compatibility guidance
Proof expectations:
- downstream repo evidence, not only repo-local docs
- docs and issue surfaces aligned with what the canaries actually show
Out of scope:
- turning every downstream repo into the same workflow shape at once
- treating compatibility forges as equal-maturity product surfaces
Workstream C: Builder And Publication Boundary
Primary owner issue:
#215
Scope:
- narrow FlakeHub RFC
- builder/publication wording cleanup
- Bazel identity follow-through where needed
Expected deliverables:
- one small RFC-style decision summary
- one clean public explanation of acceleration versus publication
- supporting doc cleanup where FlakeHub wording still leaks into runtime
Proof expectations:
- every FlakeHub-facing statement stays consistent with:
- workflow-owned Nix bootstrap
- ARC horizontal scaling
- Attic/Bazel as acceleration
Out of scope:
- making FlakeHub part of the live runtime contract
- using FlakeHub as the fix for runner sizing or bootstrap
Workstream D: Metrics And Long-Tail Platform Lanes
Primary owner issues:
#212#66#44
Scope:
- benchmark methodology
- package-management RFC work
- GPU runner docs and later runner classes
Expected deliverables:
- benchmark scorecard and method
- scoped RFC outputs for the non-runner lanes
Proof expectations:
- concrete methodology and comparison axes
- no vague “competitiveness” notes without measurement structure
Out of scope:
- distracting from the active runtime and adoption cleanup lanes
Merge And Review Order
Preferred execution order for new cleanup work:
- runtime integrity slices that unblock or validate
honey - downstream-contract slices that use live canary evidence
- publication-boundary slices that summarize and narrow, not expand
- benchmarking and longer-tail slices
If a change touches multiple workstreams, it should usually be split unless the shared code or docs are inseparable.
Canonical Outputs Per Workstream
These are the surfaces that should stay authoritative during cleanup:
- Workstream A:
current-state.md,roadmap.md,gloriousflywheel-builder-runtime-sprint-priorities-2026-04-16.md - Workstream B:
cleanup-program.md,gloriousflywheel-downstream-adoption-and-migration-kit-2026-04-16.md - Workstream C:
rfcs.md,architecture/bzlmod-topology.md - Workstream D:
gloriousflywheel-runner-benchmark-methodology-2026-04-16.md
Other notes may support these, but should not quietly replace them.
Explicit Non-Goals For This Cleanup Cycle
Do not let this cleanup program drift into:
- a fresh milestone taxonomy exercise
- a new all-at-once runtime refactor
- a public promise that GitLab, Codeberg, and GitHub are equally mature
- FlakeHub-as-runtime or FlakeHub-as-bootstrap language
- speculative storage or MinIO work without a separately scoped lane
Remaining Local-Only Paths
The repo still has intentionally local or not-yet-scoped paths outside this cleanup structure:
config/backends/tofu/modules/minio-operator/tofu/modules/minio-tenant/
Those should not be silently folded into unrelated cleanup PRs.
Recommended Next Review
When reassessing cleanup progress, ask:
- which workstream moved
- what live proof changed
- which canonical surface was updated
- whether the change reduced ambiguity or only moved text around