fix: Add GCP deployment information to Experimental section (#3219)

* fix: Add GCP deployment information to Experimental section

* fix: removed unnecessary lines

Author: digrajpu

docs/experiments.md

@@ -6,6 +6,17 @@ SC4S can now be deployed on AWS ECS/Fargate and Google Kubernetes Engine (GKE).
 See the [ECS/Fargate getting started guide](gettingstarted/ecs-fargate.md) and [GKE getting started guide](gettingstarted/gke.md) for the task definition and setup steps.
 These guides are in an experimental phase now, and should be treated as such - it's likely they will be changed / improved in the future.
 
+### GCP (Google Cloud Platform)
+SC4S can now be deployed on Google Cloud Platform (GCP) using Google Kubernetes Engine (GKE).
+See the [GKE getting started guide](gettingstarted/gke.md) for setup steps.
+This guide is in an experimental phase and should be treated as such - it's likely it will be changed / improved in the future.
+
+**GCP-specific notes:**
+
+- GCP automatically provisions a Passthrough Network Load Balancer (L4) when `service.type: LoadBalancer` is set
+- Active TCP connections may be reset during HPA scale-up and scale-down as the GCP LB redistributes connections to new pods. SC4S disk buffer (`persistence.enabled: true`) prevents event loss during these resets.
+- HPA requires the GKE Node Autoscaler to be enabled. Without it, new pods stay `Pending` indefinitely when all nodes are occupied. See the [GKE getting started guide](gettingstarted/gke.md) for details.
+
 ## > 3.12.0
 `SC4S_USE_NAME_CACHE=yes` supports IPv6.