Scale biz_state collects with dedicated workers and non-blocking UI poll.

Add PG claim/NE mutex, persist pool, and biz_state_worker replicas; fix double-SSH and row-count bugs; stop 32m collectNow while-loop from freezing page switches.

Co-authored-by: Cursor <cursoragent@cursor.com>

.env.example

@@ -66,6 +66,15 @@ NETX_UME_NOTIFICATION_TOPIC=ALARM
 # (start_netx.ps1/.sh do this automatically). Worker writes heartbeat for API /metrics.
 # NETX_RUN_INLINE_SCHEDULERS=false
 # NETX_SCHEDULER_HEARTBEAT_PATH=data/runtime/scheduler_heartbeat.json
+# Biz-state dedicated workers (default on with split start): claim queued batches via PG.
+# Rule of thumb: NETX_BIZ_STATE_MAX_CONCURRENT_TASKS * 2 <= NETX_CLI_MAX_CONCURRENT
+# NETX_BIZ_STATE_DEDICATED_WORKERS=true
+# NETX_BIZ_STATE_MAX_CONCURRENT_TASKS=16
+# NETX_BIZ_STATE_WORKER_COLLECT_THREADS=8
+# NETX_BIZ_STATE_PERSIST_WORKERS=4
+# NETX_BIZ_STATE_WORKER_REPLICAS=2
+# NETX_BIZ_STATE_SPOOL_DIR=data/biz_state_spool
+# NETX_BIZ_STATE_PERSIST_EVERY_CMDS=8
 # --- Multi-user shared-server capacity (defaults in Settings already match these) ---
 # NETX_DB_POOL_SIZE=40
 # NETX_DB_MAX_OVERFLOW=40
@@ -94,6 +103,6 @@ NETX_UME_NOTIFICATION_TOPIC=ALARM
 # NETX_NE_COLLECTION_KEEP_DAYS=14
 # Heavier fleets: raise CLI/DB together; also ensure Postgres max_connections and bastion session limits.
 # One-click start (recommended): .\scripts\start_netx.ps1 -Background -WithWeb
-#   -> API with NETX_RUN_INLINE_SCHEDULERS=false + auto-started netx_api.worker + optional Vite
+#   -> API + netx_api.worker + N× netx_api.biz_state_worker + optional Vite
 # Legacy single-process: add -InlineSchedulers
 # Stop all: .\scripts\stop_netx.ps1