Clarify Redis session routing is conditional on Redis being configured

docs/toolhive/guides-k8s/run-mcp-k8s.mdx

@@ -453,6 +453,26 @@ The proxy runner handles authentication, MCP protocol framing, and session
 management; it is stateless with respect to tool execution. The backend runs the
 actual MCP server and executes tools.
 
+### Session routing for backend replicas
+
+For stateful MCP servers, once a client establishes a session with a specific
+backend pod, subsequent requests in that session need to reach the same pod.
+When `backendReplicas > 1` and Redis session storage is configured, the proxy
+runner uses Redis to store a session-to-pod mapping so every proxy runner
+replica knows which backend pod owns each session. Stateless backends can use
+non-sticky routing by setting `sessionAffinity` to `None`.
+
+When Redis session storage is configured and a backend pod is restarted or
+replaced, its entry in the Redis routing table is invalidated and the next
+request reconnects to an available pod — sessions are not automatically migrated
+between pods.
+
+Without Redis session storage, the proxy runner relies on Kubernetes `ClientIP`
+session affinity on the backend Service. `ClientIP` affinity is unreliable
+behind NAT or shared egress IPs, and there is no shared session-to-pod mapping,
+so a pod restart or replacement can silently misroute subsequent requests to the
+wrong pod.
+
 Common configurations:
 
 - **Scale only the proxy** (`replicas: N`, omit `backendReplicas`): useful when
@@ -486,9 +506,12 @@ replica management to an HPA or other external controller.
 
 :::note
 
-The `SessionStorageWarning` condition fires only when `spec.replicas > 1`.
-Scaling only the backend (`backendReplicas > 1`) does not trigger a warning, but
-backend client-IP affinity is still unreliable behind NAT or shared egress IPs.
+The `SessionStorageWarning` condition fires only when `spec.replicas > 1`
+(multiple proxy runner pods). It does not fire when only `backendReplicas > 1`,
+but `sessionAffinity: ClientIP` is unreliable behind NAT or shared egress IPs.
+For stateful workloads or when per-session routing must remain consistent across
+backend pods, Redis session storage is strongly recommended when
+`backendReplicas > 1`. Stateless workloads can use `sessionAffinity: None`.
 
 :::