Select Field - group options (optgroup)

[bratvanov]

Select fields in HTML use the optgroup element to visually group options.

React Select's Group and GroupHeading components replicate this functionality and are utilized in Payload's Relationship Field to display relationTo slugs as group headings.

It would be a valuable enhancement if Payload's Select field also supported option grouping. While this likely wouldn't impact stored data, it would greatly improve UI/UX by organizing large lists of options more effectively.

Similarly, the Payload Form Builder plugin should also support option groups in select fields.

[seunghyukchoe]

Cycle 1 complete — backlog cleared, zero open PRs

main is at 97b56ce with 1,557 tests passing (1,494 at cycle open).

Merged this cycle

PR	What landed
#79	Typed refusal for non-orthogonal saved footprint poses — a verified untyped crash through the placement boundary
#57	Atomic route-bundle preview (all-or-nothing multi-net plans, ADR-0066)
#58	Bounded multilayer proposals, 2–8 signal layers (ADR-0068)
#59	Route-aware placement scoring, opt-in and advisory (ADR-0067)
#81	Repository professionalization: docs IA, ADR index, link checker

Closed as superseded with evidence: #48, #49, #50 — each proven by byte-identical blobs or commit containment on main, each spun out into follow-up issues rather than dropped.

What the review layer caught

Every substantive PR had a real defect the green suite did not: a schema-valid request that crashed untyped; an unrelated change that silently reversed security decision SEC-113 (DRC output moved from DEVNULL to never-read files); a silent break of two-layer candidate identity while the ADR, ledger, changelog and PR body all claimed identity was preserved; and a benchmark measuring a different thing than it claimed (the compared candidate sets were disjoint, not re-ranked).

One finding is a negative result worth stating plainly: at the honest probe count, route-aware scoring produces a worse placement than the baseline (391 mm vs 359 mm total probe length), and all four "unrouted" probes turned out to be off_grid refusals rather than routability statements. That is recorded in B-082 and ADR-0067 rather than buried; whether it should gate integration is a maintainer call.

Structural issues opened

#80 (no golden identity pins anywhere — how the identity break passed 1,506 tests), #82 (check_ledgers.py never inspects IDs), #83 (ADR/ledger IDs collide across concurrent branches — observed for real when two branches both claimed ADR-0066), plus #72/#74 (courtyard fidelity and ring nesting) and #75–#78 (live IPC hardening).

Agentic hours

Finished this cycle: ~26 agent-hours — 5 deep PR reviews with local real-KiCad gates, 5 remediation passes, 1 professionalization pass, 4 merge-conflict resolutions including an ADR renumbering.

Remaining to finish the roadmap: ~70–95 agent-hours. M1 ~8–14, M2 ~15–26 (negotiated congestion and fill-aware routing are the big items), M3 ~8–14, M4 ~12–24, M5 ~13–30 (Rust/GPU/physics remains the most speculative). Three follow-up PRs are already in flight against that total.