2.0.0-beta.19 | `createOptimisticStore` a derived optimistic store in another optimistic store does not "inherit" the pending when using `affects()`

[mizulu]

Describe the bug

s2.a does not show isPending even though it is derived from s.a which is pending by an explicit call to affects(s,'a')

  let [s,ss] = createOptimisticStore(()=>{
    return {a:1}
  }, {a:0})

  let [s2,ss2] = createOptimisticStore(()=>{
    return s 
  }, {a:0} as unknown as any)


  console.log('s.a',isPending(()=>s.a))
  console.log('s2.a', isPending(()=>s2.a))
  affects(s,'a')
  console.log('affects (s,"a")')
  console.log('s.a',isPending(()=>s.a))  // s.a is pending due to an explict affect 
  console.log('s2.a', isPending(()=>s2.a)) // s2.a is derived from s.a but is not pending 

Your Example Website or App

https://s.olid.uk/id/rp_IuUpuTqO9bYy85uoFGA

Steps to Reproduce the Bug or Issue

run code.

Expected behavior

derived key in an optimistic store, from another optimistic store, should also derive the explicit pending set by the affects() call on the store key.

Screenshots or Videos

Platform

.

Additional context

source for llm

[ryansolid]

Fixed in e156386 on next. Root cause: your second store's projection lands the first store's value, and reconcile adopts the source's raw object as the derived store's backing — so reads through s2 never touch s's node map, where a keyed affects(s, "a") mark lived. Keyless marks already resolved by raw identity (#2882); keyed marks now register an identity scope too (owning record's raw, narrowed to the key), so isPending(() => s2.a) witnesses the mark through the shared backing record, stays scoped to the marked key (s2.b stays crisp), and releases with the transaction as usual. Spec pin added under A24; regression tests in tests/question-scoped-pending.test.ts ("cross-family raw sharing"). Thanks for the tidy repro!

[mizulu]

confirmed
https://s.olid.uk/id/Hoz6ICr5R3aASEvAMFOmEQ @ d2f83c0