diff --git a/docker-compose.dev-sx-host.yml b/docker-compose.dev-sx-host.yml
index 63536980..0e30e483 100644
--- a/docker-compose.dev-sx-host.yml
+++ b/docker-compose.dev-sx-host.yml
@@ -29,6 +29,10 @@ services:
       SX_ADMIN_USER: admin
       SX_ADMIN_PASSWORD: "sx-host-camper-van-2026"
       SX_SESSION_SECRET: "ra-host-sess-7c1f9b3e2a8d4056"
+      # Serving-mode JIT: bytecode-compile hot SX (esp. the Datalog/relations path)
+      # on the epoch serving channel. Validated: host conformance 271/271 under JIT,
+      # 5.4x faster (1m43s -> 19s). Default-OFF gate, opt in here.
+      SX_SERVING_JIT: "1"
       OCAMLRUNPARAM: "b"
     volumes:
       # SX source (hot-reload on container restart)
diff --git a/lib/host/serve.sh b/lib/host/serve.sh
index b2a1090d..3a972938 100755
--- a/lib/host/serve.sh
+++ b/lib/host/serve.sh
@@ -99,6 +99,17 @@ EPOCH=1
   for M in "${MODULES[@]}"; do
     echo "(epoch $EPOCH)"; echo "(load \"$M\")"; EPOCH=$((EPOCH+1))
   done
+  # Serving-mode JIT: exclude the request-path SX (host app + Dream framework) from
+  # JIT. Those functions miscompile on first call in the http-listen + cek_run_with_io
+  # path (a map/rest arg bug → 500, then self-heals via CEK fallback — but the first
+  # hit fails). They're IO-bound anyway, so CEK is no slower. The JIT win is the
+  # Datalog/relations saturation (dl-*/relations-*), which JITs cleanly and stays on
+  # (e.g. /tags 2.5s -> 0.76s).
+  if [ "${SX_SERVING_JIT:-}" = "1" ]; then
+    echo "(epoch $EPOCH)"
+    echo "(eval \"(jit-exclude! \\\"host/*\\\" \\\"dream-*\\\" \\\"dr/*\\\")\")"
+    EPOCH=$((EPOCH+1))
+  fi
   # Point the blog at the DURABLE file backend (persists under $SX_PERSIST_DIR),
   # then idempotently seed a welcome post (sx_content = SX element markup, the
   # editor's content model). Re-seeding is a no-op if the slug already exists.