fix(db): preserve deferred demand and prepare SSR snapshots

.agents/skills/oracle-authoring/SKILL.md

@@ -19,21 +19,35 @@ first. Auditors check the order of your work and penalize shortcuts.
 Work in this order when a production change is part of the task:
 
 1. State the law and its authority.
-2. Extend the history grammar, driver, or observation until generated
+2. Inventory the materially affected public receiving paths and their promised
+   observation cuts. Include required real-host receiving witnesses in this
+   pre-edit gate; a controlled core witness does not supply their evidence.
+3. Extend the history grammar, driver, or observation until generated
    histories reach the state where the law matters. Declare that state with
-   `createOracleReach` (ORC-015).
-3. Run the extended oracle on the unchanged production code and record the RED
-   failure.
-4. Change production, with the oracle running after each attempt.
-5. Record the GREEN result and a hostile mutant that fails.
+   `createOracleReach` (ORC-015) for each required receiving path.
+4. Run those comparisons on unchanged production. Record the defect's RED
+   result and the neighboring guarantees' results. A setup failure or missing
+   required witness blocks production edits; repair the test path first.
+5. Change production, with the oracle running after each attempt.
+6. Record the GREEN result and a hostile mutant that fails.
 
-Do not change production before step 3. If a model rule disagrees with
+Do not change production before step 4. Retrospective runs or a documented
+process gap do not satisfy this gate. If a model rule disagrees with
 production, find its authority before you edit either one. A rule that you
 change only so that it matches production output is model hacking.
 
-Read [the oracle guide](../../../docs/contributing/oracle-rubric.md),
-[the glossary](../../../docs/contributing/glossary.md), and the relevant owner
-and limits in [the coverage map](../../../docs/contributing/oracle-coverage.md).
+Read [the oracle guide](../../../docs/contributing/oracle-rubric.md) and
+[the glossary](../../../docs/contributing/glossary.md) in full, then read the
+relevant owner and limits in
+[the coverage map](../../../docs/contributing/oracle-coverage.md).
+
+After a handoff or compaction, reopen required sources whose relevant contents
+are absent from the current context. A prior agent's statement that it read
+them does not supply their definitions. Apply the glossary's model alignment
+rule before naming or renaming a state, action, boundary, or observation. Use
+canonical terms and declare abstractions that combine or split production
+concepts. Delegated agents must do the reading relevant to their own scope.
+
 Follow the subsystem reading requirements in AGENTS.md.
 
 Keep unit tests and standalone concrete type assertions out of oracle files

.agents/skills/oracle-review/SKILL.md

@@ -9,9 +9,18 @@ Ask both whether the oracle enforces its stated law and whether that is the
 right law to promise. A passing test can coexist with a wrong theory; a failing
 test can coexist with correct production.
 
-Read [the oracle guide](../../../docs/contributing/oracle-rubric.md),
-[the glossary](../../../docs/contributing/glossary.md), and the relevant owner
-and limits in [the coverage map](../../../docs/contributing/oracle-coverage.md).
+Read [the oracle guide](../../../docs/contributing/oracle-rubric.md) and
+[the glossary](../../../docs/contributing/glossary.md) in full, then read the
+relevant owner and limits in
+[the coverage map](../../../docs/contributing/oracle-coverage.md).
+
+After a handoff or compaction, reopen required sources whose relevant contents
+are absent from the current context. A prior agent's statement that it read
+them does not supply their definitions. Apply the glossary's model alignment
+rule before naming or renaming a state, action, boundary, or observation. Use
+canonical terms and declare abstractions that combine or split production
+concepts. Delegated agents must do the reading relevant to their own scope.
+
 Follow the subsystem reading requirements in AGENTS.md. Guide conformance is
 judged by its ORC requirements, not by use of these instruments or headings.
 
@@ -59,6 +68,10 @@ Report each one as a finding:
 
 - A production change without a RED result from the extended oracle on the
   parent commit.
+- Required public receiving paths tested only after the production repair,
+  or aggregate reach that hides an unexercised receiving path. Check the
+  pre-edit inventory, actual comparison results, and their order. Retrospective
+  evidence or documentation of a gap does not satisfy the pre-edit gate.
 - A model rule that changed to match production output without an authority.
 - A claim about a state that no declared reach shows generated histories enter.
 - A skipped, narrowed, or relaxed comparison without a coverage-map entry and a

.changeset/resume-independent-subscription-demand.md

@@ -0,0 +1,18 @@
+---
+'@tanstack/db': patch
+'@tanstack/react-db': patch
+'@tanstack/svelte-db': patch
+'@tanstack/react-router-with-db': patch
+---
+
+Start the required Collection sync run when an ordinary subscriber follows an abandoned suspended render. Deferred acquisition still starts no idle source sync run or acquisition attempt.
+
+Ordinary subscription admission and deferred acquisition resumption throw during cleanup, including automatic GC. Effects created then return disposed handles. Failed setup releases partial ownership, and throwing callbacks cannot prevent healthy observers receiving cleanup status. Subscribers surviving cleanup retain demand for an explicit restart.
+
+SSR captures each reader's rows and Collection status. Hydration preserves that view through browser preloads, runtime errors, sibling commits, and delayed streamed records. Missing records reject when transport ends. Known pending results survive normal record-stream close. A new stream after client cleanup starts independently of prior failures.
+
+Empty loading SSR needs no preload. Rendering rows or readiness requires matching query preparation or explicit source Collection preparation. Streaming Suspense prepares its result. Unsupported preparation throws with guidance. Prepared non-streaming Suspense remains supported. The DbClient query hash includes an infinite query's initial limit and offset.
+
+Apps without `DbClient` exclude its SSR implementation. Its API stays unchanged.
+
+Svelte roots initialized with `hydrate()` must set `hydrating` on `DbProvider`. Nested providers inherit it. Each provider clears initial hydration after mount so new children read live rows.

.github/workflows/pr.yml

@@ -45,9 +45,11 @@ jobs:
       - name: Check private-member mangle cache
         run: pnpm run check:mangle
       - name: Build DB packages
-        run: pnpm --filter @tanstack/db-ivm --filter @tanstack/db build
+        run: pnpm --filter @tanstack/db-ivm --filter @tanstack/db --filter @tanstack/react-db build
       - name: Check minified public API
         run: pnpm run test:minified-db
+      - name: Check client-only query bundles
+        run: pnpm run test:ssr-bundle
       - name: Check production error text
         run: pnpm run test:production-errors
       - name: Check DB tests against built dist

docs/contributing/glossary.md

@@ -210,6 +210,8 @@ Use nouns for state and verbs for transitions:
 - A sync transaction **applies** when its writes and events become visible.
 - A replay or repair **publishes** one coherent public snapshot.
 - Cleanup start **invalidates** a sync run and **closes** restart admission.
+  Ordinary subscription admission and deferred acquisition resumption **throw**
+  during active cleanup without changing subscriber ownership.
   Cleanup **settles** after adapter cleanup settles and the Collection publishes
   `cleaned-up`. Restart then **starts** a new sync run.
 - Truncate replay **replaces** source state inside the current sync run and