perf(app): restore live sessions immediately after resume

Foregrounding now reconnects every host without waiting for scheduled backoff, while cached turns, timelines, agents, and workspaces remain internally consistent until authoritative catch-up arrives. Replica persistence is keyed and incremental so ordinary updates avoid whole-directory serialization.

docs/architecture.md

@@ -104,7 +104,7 @@ Cross-platform React Native app that connects to one or more daemons.
 
 - Expo Router navigation (`/h/[serverId]/workspace/[workspaceId]`, `/h/[serverId]/agent/[agentId]`, etc.). The `workspaceId` URL segment is an opaque workspace id, not a directly meaningful filesystem path.
 - `HostRuntimeController` manages saved host connections, reconnection, and per-host runtime state. Direct TCP and relay connections use the ordinary client transport; desktop socket, pipe, and SSH connections cross one Electron-owned transport boundary. SSH only tunnels to an already-running daemon.
-- `runtime/replica-cache` is typed storage behind the directory and timeline owners. It never observes or mutates `SessionStore`.
+- `runtime/client-replica` owns typed storage behind narrow host-lifecycle, directory, and timeline capabilities. Its cache internals never escape the module or observe `SessionStore`.
 - `runtime/directory-sync` owns directory cache selection and network reconciliation. On demand it paints accepted rows for one host, then passes the persisted per-entity cursor through `project.list`, `fetch_workspaces`, and `fetch_agents`; the daemon returns each entity's latest projection when its sequence is newer, plus tombstones.
 - `workspace-labels` owns one sequenced catalog replica per connected host, the deterministic cross-host projection that surfaces spanning hosts use (the filter page, the manager), and the per-host resolution a workspace row's chips use. Two hosts may give one name different colors, so a row resolves against its own host's catalog and a merged answer would be wrong there. Catalogs never synchronize between hosts; assignment creates a missing definition only on the target host. On the daemon, catalog and assignment rewrites share a journaled commit boundary. Startup recovery completes that commit before workspace or catalog publication.
 - `SessionContext` wraps the daemon client for the active session

docs/data-model.md

@@ -574,8 +574,8 @@ Right-sidebar client state splits on whether it is determined by the directory o
 
 The durable client replica uses IndexedDB on browser/Electron and expo-sqlite on native. Rows use the
 compound key `(serverId, kind, id)`; kinds are `agent`, `workspace`, `project`, `timeline`, and
-`checkpoint`. Directory entities have individual rows. Timeline and checkpoint use the singleton id
-and have at most one row per host.
+`checkpoint`. Directory entities have individual rows, timelines use the agent id, and the checkpoint
+uses the singleton id.
 
 The store is a typed persistence boundary. It returns values to directory and timeline owners and
 accepts their explicit commits; it never reads or writes UI state. Reads are scoped to the requested
@@ -585,6 +585,10 @@ An invalid directory row and its affected checkpoint cursor are repaired in one
 later launch cannot accept a checkpoint for a partial baseline. Directory changes and their
 checkpoint are also applied in one transaction.
 
+Owners submit keyed values and tombstones, not directory snapshots. The cache coalesces each key while
+the value remains structured, then serializes only the surviving changed rows during the deferred
+write. App backgrounding flushes immediately; the ordinary write window is one second.
+
 The cache is capped at 32 MiB and evicts whole hosts in least-recently-written order. Budget
 bookkeeping may scan opaque row sizes during a deferred write, never during host registry startup or
 before a requested cache row can paint. The row store is not encrypted. A cached timeline can contain

docs/timeline-sync.md

@@ -152,7 +152,7 @@ remaining page through the existing stream reducer. It must not append full proj
 live prefix.
 
 Every path that sends a message to an agent — composer send, dictation accept-and-send, queued
-send-now, and the automatic queue drain in `HostRuntime` — goes through
+send-now, and the session-data owner's automatic queue drain — goes through
 `dispatchComposerAgentMessage` with a submission writer. There is no second transport for the same
 product action: calling `client.sendAgentMessage` directly skips the submitted row and the pending
 footer, and permanently drops attachments because the daemon does not echo them back.
@@ -196,9 +196,9 @@ fork positions retain the canonical `turnId` boundaries.
 The compatibility boundary for older daemons is snapshot normalization: running/idle status becomes an
 anonymous active turn or idle state once, and downstream code consumes the same activity shape. The app
 does not combine anonymous lifecycle events, timestamps, timeline rows, and resume coverage to infer a
-second running state. Disconnect and replica removal remain destructive close boundaries. Elapsed time
-comes only from turn liveness, never from submission records or whichever timeline rows happen to be
-mounted.
+second running state. Disconnect preserves the last replicated turn until cache or network hydration
+advances it; replica removal remains the destructive close boundary. Elapsed time comes only from turn
+liveness, never from submission records or whichever timeline rows happen to be mounted.
 
 The daemon records one canonical submitted user row at acceptance. Its wire `messageId` is the
 submission's `clientMessageId`, so the row is born with its final identity and remains immutable on