Slate Core Deleted Test-Family Closure Consensus Plan

RALPLAN-DR Summary

Principles

1. restore representative Slate value, not dead legacy bulk
2. every surviving gap must become an explicit better-cut
3. keep the current engine and proof stack stable while widening coverage
4. prefer code-bearing contract suites over doc-only confidence

Decision Drivers

1. packages/slate/test/** is still the biggest deleted family bucket left
2. maintainers need evidence for what was recovered vs intentionally cut
3. the current restored slices already show a workable pattern: representative family contracts plus real package verification

Viable Options

Option A: Keep restoring representative family contracts cluster by cluster

Pros:

• real package work
• keeps the proof stack readable
• scales better than resurrecting hundreds of tiny legacy fixtures

Cons:

• requires judgment about what is representative
• some maintainers may still ask for more raw legacy visibility

Option B: Bulk restore the deleted fixture tree as-is

Pros:

• maximum legacy visibility
• fewer decisions per file up front

Cons:

• huge volume, noisy maintenance cost
• reintroduces many rows that no longer map cleanly to the rewritten engine
• high risk of spending time proving dead seams

Option C: Stop restoring and better-cut the rest immediately

Pros:

• fastest path to a smaller surface
• avoids endless test-port work

Cons:

• too early while the maintainers-facing deletion bucket is still huge
• weak evidence if we cut families we have not sampled properly

Chosen Option

• Option A

Why chosen:

• it is the best balance between real recovery and honest scoping

Task Statement

Close the remaining packages/slate/test/** deleted-family bucket by mapping every deleted row to current proof, restoring the uncovered current-value clusters, then converting the incompatible residue into explicit skip entries with better-cut rationale.

Scope

In Scope

• deleted packages/slate/test/** family closure
• representative contract suites for the still-open families
• any package code fixes needed to make those restored suites pass
• proof-ledger and file-review-ledger updates for each recovered family
• explicit better-cut decisions for families we decide not to restore further

Out of Scope

• packages/slate-react/test/** deleted-family closure
• packages/slate-history/test/** deleted-family closure
• source-bucket restore-or-cut under packages/slate/src/**
• runtime quarantine work

Decision

Treat deleted packages/slate/test/** closure as an uncovered-proof-delta problem, not a family-ordering ritual.

Start with a classification pass for every remaining family:

1. mirrored now
   • already covered by snapshot-contract.ts or an existing contract file
2. supportable next
   • uncovered rows that still map cleanly to the current engine and public contract
3. explicit skip
   • legacy rows that no longer belong and should be called out directly

Use explicit skip as the accounting label and better-cut rationale as the explanation attached to that label.

Then execute the remaining work in uncovered-delta order, not family-label order:

1. classify the family
2. restore the highest-value uncovered cluster
3. extend an existing contract when that keeps proof closest to the live API
4. create a new contract file only when it isolates a real claim better
5. fix package code only if the restored proof exposes a real gap
6. rerun:
   • restored family tests
   • yarn test:mocha
   • yarn workspace slate-react run test
   • yarn workspace slate-dom test
   • yarn test:custom
   • yarn lint:typescript
7. update the proof + file-review ledgers
8. close the family with an explicit note:
   • uncovered public-value delta closed
   • remaining rows explicitly skipped
   • ledger reflects the final read

Execution Model

Phase 0: Classification Matrix

Goal:

• publish an auditable matrix for every still-open deleted family: mirrored now, supportable next, explicit skip

Families:

• transforms/**
• remaining deleted rows within interfaces/** after excluding claims already mirrored by current proof
• remaining deleted rows within operations/** after excluding claims already mirrored by current proof
• remaining deleted rows within normalization/** after excluding claims already mirrored by current proof
• remaining deleted rows within utils/** after excluding claims already mirrored by current proof
• index.js
• jsx.d.ts

Output:

• one written Family Closure Matrix inside the execution artifact with these columns:
  • deleted path or cluster id
  • family
  • deleted count
  • legacy claim
  • current proof owner file#line
  • status
  • rationale
  • new proof file if any
  • ledger row touched
• the matrix totals must reconcile exactly to the parent audit counts:
  • interfaces/** = 576
  • transforms/** = 408
  • operations/** = 31
  • normalization/** = 20
  • utils/** = 11
  • index.js = 1
  • jsx.d.ts = 1
  • total packages/slate/test/** deleted rows = 1048
• no restore work starts before this table exists

Phase 1: Largest Uncovered Delta First

Goal:

• start with the family whose matrix shows the highest uncovered deleted count still marked supportable next

Important read:

• do not blindly start with transforms/**
• do not blindly assume interfaces/** is next
• snapshot-contract.ts already mirrors a large amount of transform and value semantics, so duplicated proof is failure, not progress

Exit:

• first uncovered high-value cluster is restored without duplicating existing snapshot proof

Phase 2: Family-by-Family Delta Closure

Goal:

• keep restoring only uncovered supportable clusters until each family is exhaustively classified and honestly closed

Restore targets may land as:

• extensions to snapshot-contract.ts
• extensions to an existing contract file
• a new dedicated contract file when isolation is clearer than extension

Exit:

• every remaining family is classified
• every supportable uncovered cluster is either restored or reclassified to explicit skip with a stated incompatibility reason

Phase 3: Entry-Point Residue Decisions

Goal:

• decide packages/slate/test/index.js and packages/slate/test/jsx.d.ts explicitly instead of leaving them as vague residue

Exit:

• each entrypoint is restored if still needed
• otherwise it is explicitly skipped in the ledgers and plan notes with better-cut rationale

Phase 4: Closeout Notes

Goal:

• produce a maintainer-auditable closeout note per family

Each note must say:

• what was already mirrored
• what was restored
• what was explicitly skipped
• why no more restore work is needed

Family Closure Matrix

Before any new restore lands, fill this matrix and keep it current.

deleted path or cluster id	family	deleted count	legacy claim	current proof owner file#line	status (mirrored now / supportable next / explicit skip)	rationale	new proof file if any	ledger row touched
packages/slate/test/interfaces/**:cluster-*	interfaces/**	576 total	value/query/runtime-guard rows	snapshot-contract.ts, interfaces-contract.ts	mirrored now + explicit skip	helper-heavy surface is already mirrored by current proof; only deleted CustomTypes declaration-merging rows are explicit skip	2026-04-09-slate-v2-interfaces-family-deleted-test-closure.md	major file/test deletion review
packages/slate/test/transforms/**:cluster-*	transforms/**	408 total	helper transform rows	snapshot-contract.ts, clipboard-contract.ts, transforms-contract.ts	mirrored now + recovered now + explicit skip	current narrow transform contract is either already mirrored, directly recovered, or explicitly outside the live claim	2026-04-09-slate-v2-transforms-family-deleted-test-closure.md	major file/test deletion review
packages/slate/test/operations/**:cluster-*	operations/**	31 total	raw operation rows	operations-contract.ts	recovered now + explicit skip	current raw-operation seam is directly recovered; legacy custom-selection and null/undefined removal-sentinel rows are explicit skip	2026-04-09-slate-v2-operations-family-deleted-test-closure.md	major file/test deletion review
packages/slate/test/normalization/**:cluster-*	normalization/**	20 total	normalization rows	normalization-contract.ts	mirrored now	current default-vs-explicit normalization split mirrors the deleted normalization rows without reopening broader always-on coercion	2026-04-09-slate-v2-normalization-family-deleted-test-closure.md	major file/test deletion review
packages/slate/test/utils/**:cluster-*	utils/**	11 total	helper rows	text-units-contract.ts	mirrored now + explicit skip	surviving string-unit value is mirrored; deleted deep-equal/** rows are explicit skip	2026-04-09-slate-v2-utils-and-root-test-entrypoint-closure.md	major file/test deletion review
packages/slate/test/index.js	index.js	1	test entrypoint	none	explicit skip	old fixture-harness root entrypoint is replaced by direct contract files plus package-local Mocha globs	2026-04-09-slate-v2-utils-and-root-test-entrypoint-closure.md	major file/test deletion review
packages/slate/test/jsx.d.ts	jsx.d.ts	1	test typing entrypoint	none	explicit skip	old JSX typing shim belonged to the deleted fixture harness, not the current direct-contract lane	2026-04-09-slate-v2-utils-and-root-test-entrypoint-closure.md	major file/test deletion review

Closure rule:

• no family is done until its matrix rows reconcile to the audit counts exactly
• no new proof lands unless the matrix cites the existing proof it does not duplicate
• every supportable next row that is not restored must be reclassified to explicit skip with a stated incompatibility reason
• this batch closes only major file/test deletion review in release-file-review-ledger.md when all 1048 deleted packages/slate/test/** rows are mapped and every surviving uncovered public-value delta has direct proof or an explicit skip note
• this batch does not by itself close broader True Slate RC lanes

Non-Redundancy Gate

No restore may land unless the closure matrix cites the existing proof rows it does not duplicate.

Known proof owners to audit first:

• snapshot-contract.ts
• interfaces-contract.ts
• operations-contract.ts
• normalization-contract.ts
• text-units-contract.ts
• range-ref-contract.ts
• clipboard-contract.ts

Acceptance Criteria

1. Every still-open deleted packages/slate/test/** family has a written classification:
   • mirrored now
   • supportable next
   • explicit skip
2. Every cluster initially marked supportable next is either restored in code-bearing proof or reclassified to explicit skip with a stated incompatibility reason.
3. Any package-code gap exposed by restored proof is fixed in the package, not hand-waved in docs.
4. Every restored or cut family is wired into:
   • true-slate-rc-proof-ledger.md
   • release-file-review-ledger.md
5. Every family has a closeout note explaining why no more restore work is needed.
6. Batch-local verification passes after each landed family batch:
   • restored family tests
   • yarn test:mocha
   • yarn workspace slate-react run test
   • yarn workspace slate-dom test
   • yarn test:custom
   • yarn lint:typescript
7. Final closeout reruns the parent release-gate evidence or states explicitly why unchanged lanes were not reopened.

Risks

• skipping classification first can create duplicate proof and fake progress
• restoring too many literal legacy rows can waste time on dead seams
• restoring too few rows can leave maintainers unconvinced
• touching low-level transform behavior can regress the already-green snapshot / clipboard / runtime proof stack

Mitigations

• classify before restoring
• order work by uncovered proof delta, not family labels
• force matrix accounting before each restore
• extend existing proof files when that is cleaner than minting new ones
• keep the family-by-family representative strategy
• require full package verification after each family
• prefer explicit skip decisions once a family has enough representative evidence

ADR

Decision

Close the remaining deleted packages/slate/test/** bucket by restoring representative family contracts in descending value order, instead of bulk restoring the whole deleted fixture tree.

Drivers

• real code/proof progress matters more than raw restored file count
• maintainers need auditable evidence, not vibes
• the current engine already supports a lot of these families when expressed on the new seams

Alternatives Considered

• bulk restore the entire deleted fixture tree
• stop restoring and better-cut everything now

Why Chosen

• representative restores give the best signal-to-cost ratio

Consequences

• more deliberate judgment required
• lower noise than bulk restoration
• stronger maintainer story than premature cuts

Verification

Batch-local:

• restored family tests
• yarn test:mocha
• yarn workspace slate-react run test
• yarn workspace slate-dom test
• yarn test:custom
• yarn lint:typescript

Final closeout:

• rerun the parent release-gate evidence or state explicitly why unchanged lanes were not reopened
• relevant ledger/docs formatting checks