Slate v2 Built-in Normalization Recovery Lane

Goal

Recover enough of the legacy built-in normalization family to support the true Slate claim without regressing the current clipboard and range-ref contract.

Hard Constraint

Do not treat this as a quick helper port.

A naive transplant of legacy built-in normalization and fallbackElement behavior already regressed:

• mixed-inline clipboard lanes
• wrapper-list clipboard lanes
• complex fragment selection rebasing
• range-ref proof over those same families

Required Design Questions

1. Which legacy built-in constraints are still truly required for the v2 engine model?
2. Which current fragment / rebasing lanes depend on the broader tree shapes that the naive port destroyed?
3. Can fallbackElement be recovered as a scoped behavior without reintroducing blanket child-family coercion?
4. Which proofs must stay green while the lane lands?

Required Proof Before Landing

• yarn test:custom
• yarn workspace slate-react run test
• any touched clipboard / range-ref rows in:
  • packages/slate/test/clipboard-contract.ts
  • packages/slate/test/range-ref-contract.ts

Current Read

• custom normalizeNode(...) and shouldNormalize(...) are real
• safe built-in primitives now exist on the default normalizer:
  • empty-child repair
  • inline spacer insertion
  • direct-child block-only cleanup for current node-op seams
  • replace/manual-normalize block-only cleanup
• explicit-only cleanup is now a real lane:
  • Editor.normalize(...) marks the pass as explicit for heavier text cleanup
  • explicit normalize can also flatten inline-container block wrappers into inline-compatible descendants
• scoped fallbackElement recovery now exists for wrapping stray top-level or block-only text / inline children when app code delegates to the core normalizer
• broad live inline-container flattening is explicitly not part of the default live contract unless a better design survives the full clipboard / range-ref / runtime proof stack
• built-in normalization parity is still open
• blanket legacy child-family coercion is still explicitly outside the current proved surface