diff --git a/infra/default-builds/dockerfiles/test-mm-lobby-echo/Dockerfile b/infra/default-builds/dockerfiles/test-mm-lobby-echo/Dockerfile
index b052729ee..c3971456b 100644
--- a/infra/default-builds/dockerfiles/test-mm-lobby-echo/Dockerfile
+++ b/infra/default-builds/dockerfiles/test-mm-lobby-echo/Dockerfile
@@ -1,4 +1,4 @@
-FROM clux/muslrust:1.77.2 AS build
+FROM clux/muslrust:1.77.2-stable AS build
 RUN cargo new --bin /app
 WORKDIR /app
 COPY Cargo.toml ./
diff --git a/infra/tf/cloudflare_tunnels/k8s.tf b/infra/tf/cloudflare_tunnels/k8s.tf
index b18539415..3444a2dc9 100644
--- a/infra/tf/cloudflare_tunnels/k8s.tf
+++ b/infra/tf/cloudflare_tunnels/k8s.tf
@@ -47,6 +47,8 @@ resource "kubernetes_deployment" "cloudflared" {
 			}
 
 			spec {
+				priority_class_name = "service-priority"
+
 				container {
 					image = "cloudflare/cloudflared:2023.8.2"
 					name = "cloudflared"
diff --git a/infra/tf/grafana/grafana.tf b/infra/tf/grafana/grafana.tf
index 1051062dd..b363998f3 100644
--- a/infra/tf/grafana/grafana.tf
+++ b/infra/tf/grafana/grafana.tf
@@ -30,6 +30,7 @@ resource "helm_release" "grafana" {
 	chart = "grafana"
 	version = "7.3.9"
 	values = [yamlencode({
+		priorityClassName = "monitoring-priority"
 		"grafana.ini" = {
 			auth = {
 				disable_login_form = true
diff --git a/infra/tf/k8s_cluster_aws/overprovisioning.tf b/infra/tf/k8s_cluster_aws/overprovisioning.tf
new file mode 100644
index 000000000..49a0a10c5
--- /dev/null
+++ b/infra/tf/k8s_cluster_aws/overprovisioning.tf
@@ -0,0 +1,70 @@
+# Define low-priority deployments to maintain unallocated capacity on Karpenter
+# cluster to prevent premption.
+#
+# Higher replicas + lower resoruces = more smaller pods can preempt
+# Lower replicas + higher resources = larger pod can preempt, but will require bigger preempted "wedges"
+#
+# https://aws.amazon.com/blogs/containers/eliminate-kubernetes-node-scaling-lag-with-pod-priority-and-over-provisioning/
+
+resource "kubernetes_namespace" "karpenter_overprovision" {
+	metadata {
+		name = "karpenter-overprovision"
+	}
+}
+
+resource "kubernetes_priority_class" "overprovision_priority" {
+	metadata {
+		name = "overprovision-priority"
+	}
+	value = -1
+}
+
+resource "kubernetes_deployment" "overprovision" {
+	metadata {
+		name = "overprovision"
+		namespace = kubernetes_namespace.karpenter_overprovision.metadata.0.name
+		labels = {
+			app = "overprovisioning"
+		}
+	}
+
+	spec {
+		replicas = 2
+
+		selector {
+			match_labels = {
+				app = "overprovisioning"
+			}
+		}
+
+		template {
+			metadata {
+				labels = {
+					app = "overprovisioning"
+				}
+			}
+
+			spec {
+				container {
+					name = "pause"
+					image = "registry.k8s.io/pause"
+
+					resources {
+						requests = {
+							cpu	= "1"
+							memory = "500Mi"
+						}
+
+						limits = {
+							cpu = "1"
+							memory = "500Mi"
+						}
+					}
+				}
+
+				priority_class_name = kubernetes_priority_class.overprovision_priority.metadata.0.name
+			}
+		}
+	}
+}
+
diff --git a/infra/tf/k8s_infra/clickhouse.tf b/infra/tf/k8s_infra/clickhouse.tf
index 24abddea6..fd04fe7c6 100644
--- a/infra/tf/k8s_infra/clickhouse.tf
+++ b/infra/tf/k8s_infra/clickhouse.tf
@@ -27,16 +27,6 @@ resource "kubernetes_namespace" "clickhouse" {
 	}
 }
 
-resource "kubernetes_priority_class" "clickhouse_priority" {
-	count = local.clickhouse_enabled ? 1 : 0
-
-	metadata {
-		name = "clickhouse-priority"
-	}
-
-	value = 40
-}
-
 resource "helm_release" "clickhouse" {
 	count = local.clickhouse_enabled ? 1 : 0
 	depends_on = [null_resource.daemons]
@@ -57,7 +47,7 @@ resource "helm_release" "clickhouse" {
 			replicaCount = 1
 		}
 
-		priorityClassName = kubernetes_priority_class.clickhouse_priority.0.metadata.0.name
+		priorityClassName = kubernetes_priority_class.stateful_priority.metadata.0.name
 		resources = var.limit_resources ? {
 			limits = {
 				memory = "${local.service_clickhouse.resources.memory}Mi"
diff --git a/infra/tf/k8s_infra/cockroachdb.tf b/infra/tf/k8s_infra/cockroachdb.tf
index 7f48e1d59..503db6c9a 100644
--- a/infra/tf/k8s_infra/cockroachdb.tf
+++ b/infra/tf/k8s_infra/cockroachdb.tf
@@ -24,14 +24,6 @@ resource "kubernetes_namespace" "cockroachdb" {
 	}
 }
 
-resource "kubernetes_priority_class" "cockroachdb_priority" {
-	metadata {
-		name = "cockroachdb-priority"
-	}
-
-	value = 40
-}
-
 # NOTE: Helm chart is no longer supported by CockroachDB. However, it's intended to be used only for development and it's the easiest to set up.
 resource "helm_release" "cockroachdb" {
 	depends_on = [null_resource.daemons]
@@ -46,7 +38,7 @@ resource "helm_release" "cockroachdb" {
 		statefulset = {
 			replicas = local.service_cockroachdb.count
 
-			priorityClassName = kubernetes_priority_class.cockroachdb_priority.metadata.0.name
+			priorityClassName = kubernetes_priority_class.stateful_priority.metadata.0.name
 
 			resources = var.limit_resources ? {
 				limits = {
diff --git a/infra/tf/k8s_infra/imagor.tf b/infra/tf/k8s_infra/imagor.tf
index ba1cfcff5..fa3d2dbc4 100644
--- a/infra/tf/k8s_infra/imagor.tf
+++ b/infra/tf/k8s_infra/imagor.tf
@@ -56,16 +56,6 @@ resource "kubernetes_namespace" "imagor" {
 	}
 }
 
-resource "kubernetes_priority_class" "imagor_priority" {
-	count = var.imagor_enabled ? 1 : 0
-
-	metadata {
-		name = "imagor-priority"
-	}
-
-	value = 35
-}
-
 resource "kubernetes_deployment" "imagor" {
 	count = var.imagor_enabled ? 1 : 0
 	depends_on = [null_resource.daemons, module.docker_auth]
@@ -92,7 +82,7 @@ resource "kubernetes_deployment" "imagor" {
 			}
 
 			spec {
-				priority_class_name = kubernetes_priority_class.imagor_priority.0.metadata.0.name
+				priority_class_name = kubernetes_priority_class.service_priority.metadata.0.name
 				
 				# MARK: Docker auth
 				image_pull_secrets {
diff --git a/infra/tf/k8s_infra/loki.tf b/infra/tf/k8s_infra/loki.tf
index 603c9f047..556fad137 100644
--- a/infra/tf/k8s_infra/loki.tf
+++ b/infra/tf/k8s_infra/loki.tf
@@ -16,16 +16,6 @@ resource "kubernetes_namespace" "loki" {
 	}
 }
 
-resource "kubernetes_priority_class" "loki_priority" {
-	count = var.prometheus_enabled ? 1 : 0
-
-	metadata {
-		name = "loki-priority"
-	}
-
-	value = 40
-}
-
 resource "helm_release" "loki" {
 	count = var.prometheus_enabled ? 1 : 0
 
@@ -36,7 +26,7 @@ resource "helm_release" "loki" {
 	version = "5.36.0"
 	values = [yamlencode({
 		global = {
-			priorityClassName = kubernetes_priority_class.loki_priority.0.metadata.0.name
+			priorityClassName = kubernetes_priority_class.monitoring_priority.metadata.0.name
 		}
 		loki = {
 			auth_enabled = false
diff --git a/infra/tf/k8s_infra/minio.tf b/infra/tf/k8s_infra/minio.tf
index ab35ab19f..de9cf3245 100644
--- a/infra/tf/k8s_infra/minio.tf
+++ b/infra/tf/k8s_infra/minio.tf
@@ -26,14 +26,6 @@ module "minio_secrets" {
 	optional = true
 }
 
-resource "kubernetes_priority_class" "minio_priority" {
-	metadata {
-		name = "minio-priority"
-	}
-
-	value = 40
-}
-
 resource "helm_release" "minio" {
 	depends_on = [null_resource.daemons]
 	count = local.has_minio ? 1 : 0
@@ -48,7 +40,7 @@ resource "helm_release" "minio" {
 			storageClass = var.k8s_storage_class
 		}
 		replicaCount = local.service_minio.count
-		priorityClassName = kubernetes_priority_class.minio_priority.metadata.0.name
+		priorityClassName = kubernetes_priority_class.service_priority.metadata.0.name
 		resources = var.limit_resources ? {
 			limits = {
 				memory = "${local.service_minio.resources.memory}Mi"
diff --git a/infra/tf/k8s_infra/nats.tf b/infra/tf/k8s_infra/nats.tf
index fbc7d3e34..b6ca9af90 100644
--- a/infra/tf/k8s_infra/nats.tf
+++ b/infra/tf/k8s_infra/nats.tf
@@ -14,14 +14,7 @@ resource "kubernetes_namespace" "nats" {
 	}
 }
 
-resource "kubernetes_priority_class" "nats_priority" {
-	metadata {
-		name = "nats-priority"
-	}
-
-	value = 40
-}
-
+# TODO(RVTEE-105): Fix priority class
 resource "helm_release" "nats" {
 	depends_on = [null_resource.daemons]
 
@@ -37,11 +30,6 @@ resource "helm_release" "nats" {
 				replicas = local.service_nats.count
 			}
 		}
-		podTemplate = {
-			merge = {
-				priorityClassName = kubernetes_priority_class.nats_priority.metadata.0.name
-			}
-		}
 		container = {
 			env = {
 				# See https://artifacthub.io/packages/helm/grafana/grafana#nats-container-resources
diff --git a/infra/tf/k8s_infra/nomad.tf b/infra/tf/k8s_infra/nomad.tf
index 3e3d3e741..62dc8bcb3 100644
--- a/infra/tf/k8s_infra/nomad.tf
+++ b/infra/tf/k8s_infra/nomad.tf
@@ -190,16 +190,6 @@ resource "kubectl_manifest" "nomad_server_monitor" {
 	})
 }
 
-resource "kubernetes_priority_class" "nomad_priority" {
-	count = var.edge_enabled ? 1 : 0
-
-	metadata {
-		name = "nomad-priority"
-	}
-
-	value = 40
-}
-
 resource "kubernetes_stateful_set" "nomad_server" {
 	count = var.edge_enabled ? 1 : 0
 	depends_on = [null_resource.daemons]
@@ -234,7 +224,7 @@ resource "kubernetes_stateful_set" "nomad_server" {
 			}
 
 			spec {
-				priority_class_name = kubernetes_priority_class.nomad_priority.0.metadata.0.name
+				priority_class_name = kubernetes_priority_class.stateful_priority.metadata.0.name
 
 				security_context {
 					run_as_user = 0
diff --git a/infra/tf/k8s_infra/nsfw_api.tf b/infra/tf/k8s_infra/nsfw_api.tf
index 1bc2770b2..63a62dbf1 100644
--- a/infra/tf/k8s_infra/nsfw_api.tf
+++ b/infra/tf/k8s_infra/nsfw_api.tf
@@ -16,16 +16,6 @@ resource "kubernetes_namespace" "nsfw_api" {
 	}
 }
 
-resource "kubernetes_priority_class" "nsfw_api_priority" {
-	count = var.nsfw_api_enabled ? 1 : 0
-
-	metadata {
-		name = "nsfw-api-priority"
-	}
-
-	value = 40
-}
-
 resource "kubernetes_deployment" "nsfw_api" {
 	count = var.nsfw_api_enabled ? 1 : 0
 	depends_on = [null_resource.daemons, module.docker_auth]
@@ -52,7 +42,7 @@ resource "kubernetes_deployment" "nsfw_api" {
 			}
 
 			spec {
-				priority_class_name = kubernetes_priority_class.nsfw_api_priority.0.metadata.0.name
+				priority_class_name = kubernetes_priority_class.service_priority.metadata.0.name
 				
 				# MARK: Docker auth
 				image_pull_secrets {
diff --git a/infra/tf/k8s_infra/priority_class.tf b/infra/tf/k8s_infra/priority_class.tf
new file mode 100644
index 000000000..48ef405b9
--- /dev/null
+++ b/infra/tf/k8s_infra/priority_class.tf
@@ -0,0 +1,41 @@
+# Used for services that can be preempted often.
+#
+# We set almost everything to this class (even crucial infrastrcuture) because
+# we need to wait for Karpenter to boot a new node instead of shutting down
+# existing services.
+resource "kubernetes_priority_class" "service_priority" {
+
+	metadata {
+		name = "service-priority"
+	}
+	value = 50
+}
+
+# Used for anything required to monitor the other services. These should take
+# priority no matter what in order to ensure we have visibility on what's going
+# on.
+resource "kubernetes_priority_class" "monitoring_priority" {
+
+	metadata {
+		name = "monitoring-priority"
+	}
+	value = 50
+}
+
+# Used for anything stateful that should not be frequently preempted.
+resource "kubernetes_priority_class" "stateful_priority" {
+
+	metadata {
+		name = "stateful-priority"
+	}
+	value = 60
+}
+
+# Used for daemons that run on the machines and need to be scheduled no matter what.
+resource "kubernetes_priority_class" "daemon_priority" {
+
+	metadata {
+		name = "daemon-priority"
+	}
+	value = 90
+}
diff --git a/infra/tf/k8s_infra/prometheus.tf b/infra/tf/k8s_infra/prometheus.tf
index f9135dddc..2a86398dc 100644
--- a/infra/tf/k8s_infra/prometheus.tf
+++ b/infra/tf/k8s_infra/prometheus.tf
@@ -86,25 +86,6 @@ resource "kubernetes_namespace" "prometheus" {
 	}
 }
 
-# Set a high priority for Node Exporter so it can run on all nodes
-resource "kubernetes_priority_class" "node_exporter_priority" {
-	count = var.prometheus_enabled ? 1 : 0
-
-	metadata {
-		name = "node-exporter-priority"
-	}
-	value = 90
-}
-
-resource "kubernetes_priority_class" "prometheus_priority" {
-	count = var.prometheus_enabled ? 1 : 0
-
-	metadata {
-		name = "prometheus-priority"
-	}
-	value = 40
-}
-
 resource "helm_release" "prometheus" {
 	count = var.prometheus_enabled ? 1 : 0
 	depends_on = [helm_release.vpa]
@@ -122,7 +103,7 @@ resource "helm_release" "prometheus" {
 					cpu = "${local.service_node_exporter.resources.cpu}m"
 				}
 			} : null
-			priorityClassName = kubernetes_priority_class.node_exporter_priority.0.metadata.0.name
+			priorityClassName = kubernetes_priority_class.daemon_priority.metadata.0.name
 			affinity = {
 				nodeAffinity = {
 					requiredDuringSchedulingIgnoredDuringExecution = {
@@ -270,7 +251,7 @@ resource "helm_release" "prometheus" {
 					}
 				}
 			
-				priorityClassName = kubernetes_priority_class.prometheus_priority.0.metadata.0.name
+				priorityClassName = kubernetes_priority_class.monitoring_priority.metadata.0.name
 				resources = var.limit_resources ? {
 					limits = {
 						memory = "${local.service_prometheus.resources.memory}Mi"
diff --git a/infra/tf/k8s_infra/promtail.tf b/infra/tf/k8s_infra/promtail.tf
index 46bdf1945..5ce245a39 100644
--- a/infra/tf/k8s_infra/promtail.tf
+++ b/infra/tf/k8s_infra/promtail.tf
@@ -16,15 +16,6 @@ resource "kubernetes_namespace" "promtail" {
 	}
 }
 
-resource "kubernetes_priority_class" "promtail_priority" {
-	count = var.prometheus_enabled ? 1 : 0
-
-	metadata {
-		name = "promtail-priority"
-	}
-	value = 40
-}
-
 resource "helm_release" "promtail" {
 	count = var.prometheus_enabled ? 1 : 0
 
@@ -146,7 +137,7 @@ resource "helm_release" "promtail" {
 			}
 		}
 
-		priorityClassName = kubernetes_priority_class.promtail_priority.0.metadata.0.name
+		priorityClassName = kubernetes_priority_class.daemon_priority.metadata.0.name
 		resources = var.limit_resources ? {
 			limits = {
 				memory = "${local.service_promtail.resources.memory}Mi"
diff --git a/infra/tf/k8s_infra/redis.tf b/infra/tf/k8s_infra/redis.tf
index 759a62180..2003a08af 100644
--- a/infra/tf/k8s_infra/redis.tf
+++ b/infra/tf/k8s_infra/redis.tf
@@ -20,7 +20,7 @@ locals {
 	redis_node_config = {
 		for k, v in var.redis_dbs:
 		k => {
-			priorityClassName = kubernetes_priority_class.redis_priority.metadata.0.name
+			priorityClassName = kubernetes_priority_class.stateful_priority.metadata.0.name
 			resources = var.limit_resources ? {
 				limits = {
 					memory = "${local.service_redis.resources.memory}Mi"
@@ -53,13 +53,6 @@ resource "kubernetes_namespace" "redis" {
 	}
 }
 
-resource "kubernetes_priority_class" "redis_priority" {
-	metadata {
-		name = "redis-priority"
-	}
-	value = 40
-}
-
 resource "helm_release" "redis" {
 	depends_on = [null_resource.daemons]
 	for_each = local.redis_svcs
diff --git a/infra/tf/k8s_infra/traefik.tf b/infra/tf/k8s_infra/traefik.tf
index d2a4af33b..b6597b466 100644
--- a/infra/tf/k8s_infra/traefik.tf
+++ b/infra/tf/k8s_infra/traefik.tf
@@ -23,13 +23,6 @@ locals {
 	})
 }
 
-resource "kubernetes_priority_class" "traefik_priority" {
-	metadata {
-		name = "traefik-priority"
-	}
-	value = 40
-}
-
 resource "helm_release" "traefik" {
 	depends_on = [null_resource.daemons]
 
@@ -56,7 +49,7 @@ resource "helm_release" "traefik" {
 			"traefik-instance" = "main"
 		}
 
-		priorityClassName = kubernetes_priority_class.traefik_priority.metadata.0.name
+		priorityClassName = kubernetes_priority_class.service_priority.metadata.0.name
 		resources = var.limit_resources ? {
 			limits = {
 				memory = "${local.service_traefik.resources.memory}Mi"
diff --git a/infra/tf/k8s_infra/traefik_tunnel.tf b/infra/tf/k8s_infra/traefik_tunnel.tf
index 128d68b16..f179d458e 100644
--- a/infra/tf/k8s_infra/traefik_tunnel.tf
+++ b/infra/tf/k8s_infra/traefik_tunnel.tf
@@ -68,15 +68,6 @@ resource "kubernetes_namespace" "traefik_tunnel" {
 	}
 }
 
-resource "kubernetes_priority_class" "traefik_tunnel_priority" {
-	count = var.edge_enabled ? 1 : 0
-
-	metadata {
-		name = "traefik-tunnel-priority"
-	}
-	value = 40
-}
-
 resource "helm_release" "traefik_tunnel" {
 	count = var.edge_enabled ? 1 : 0
 
@@ -114,7 +105,7 @@ resource "helm_release" "traefik_tunnel" {
 			}
 		}
 
-		priorityClassName = kubernetes_priority_class.traefik_tunnel_priority.0.metadata.0.name
+		priorityClassName = kubernetes_priority_class.service_priority.metadata.0.name
 
 		tlsOptions = {
 			"ingress-tunnel" = {
diff --git a/infra/tf/k8s_infra/traffic_server.tf b/infra/tf/k8s_infra/traffic_server.tf
index 250451765..c37ae945d 100644
--- a/infra/tf/k8s_infra/traffic_server.tf
+++ b/infra/tf/k8s_infra/traffic_server.tf
@@ -107,13 +107,6 @@ resource "kubernetes_service" "traffic_server" {
 	}
 }
 
-resource "kubernetes_priority_class" "traffic_server_priority" {
-	metadata {
-		name = "traffic-server-priority"
-	}
-	value = 40
-}
-
 resource "kubernetes_stateful_set" "traffic_server" {
 	depends_on = [null_resource.daemons, module.docker_auth]
 
@@ -148,7 +141,7 @@ resource "kubernetes_stateful_set" "traffic_server" {
 			}
 
 			spec {
-				priority_class_name = kubernetes_priority_class.traffic_server_priority.metadata.0.name
+				priority_class_name = kubernetes_priority_class.service_priority.metadata.0.name
 
 				security_context {
 					run_as_user = 1000
diff --git a/infra/tf/vector/vector.tf b/infra/tf/vector/vector.tf
index a81c6cc52..e8b635cf3 100644
--- a/infra/tf/vector/vector.tf
+++ b/infra/tf/vector/vector.tf
@@ -10,13 +10,6 @@ locals {
 	clickhouse_k8s = var.clickhouse_enabled && var.clickhouse_provider == "kubernetes"
 }
 
-resource "kubernetes_priority_class" "vector_priority" {
-	metadata {
-		name = "vector-priority"
-	}
-	value = 40
-}
-
 module "secrets" {
 	source = "../modules/secrets"
 
@@ -33,7 +26,7 @@ resource "helm_release" "vector" {
 	version = "0.29.0"
 	values = [yamlencode({
 		role = "Aggregator"
-		podPriorityClassName = kubernetes_priority_class.vector_priority.metadata.0.name
+		podPriorityClassName = "service-priority"
 		resources = var.limit_resources ? {
 			limits = {
 				memory = "${local.service_vector.resources.memory}Mi"
diff --git a/lib/bolt/core/src/dep/k8s/gen.rs b/lib/bolt/core/src/dep/k8s/gen.rs
index 1b5e9a92e..ae7501160 100644
--- a/lib/bolt/core/src/dep/k8s/gen.rs
+++ b/lib/bolt/core/src/dep/k8s/gen.rs
@@ -364,18 +364,6 @@ pub async fn gen_svc(exec_ctx: &ExecServiceContext) -> Vec<serde_json::Value> {
 		}
 	});
 
-	// Create priority class
-	let priority_class_name = format!("{}-priority", service_name);
-	specs.push(json!({
-		"apiVersion": "scheduling.k8s.io/v1",
-		"kind": "PriorityClass",
-		"metadata": {
-			"name": priority_class_name,
-			"namespace": "rivet-service"
-		},
-		"value": svc_ctx.config().service.priority()
-	}));
-
 	let restart_policy = if matches!(run_context, RunContext::Test { .. }) {
 		"Never"
 	} else if spec_type == SpecType::Deployment {
@@ -395,7 +383,7 @@ pub async fn gen_svc(exec_ctx: &ExecServiceContext) -> Vec<serde_json::Value> {
 		config::ns::ClusterKind::Distributed { .. } => 30,
 	};
 	let pod_spec = json!({
-		"priorityClassName": priority_class_name,
+		"priorityClassName": "service-priority",
 		"restartPolicy": restart_policy,
 		"terminationGracePeriodSeconds": termination_grace_period,
 		"imagePullSecrets": [{
diff --git a/lib/job-runner/Dockerfile b/lib/job-runner/Dockerfile
index e29d332fa..87e9b4890 100644
--- a/lib/job-runner/Dockerfile
+++ b/lib/job-runner/Dockerfile
@@ -1,4 +1,4 @@
-FROM clux/muslrust:1.77.2
+FROM clux/muslrust:1.77.2-stable
 
 WORKDIR /app
 COPY Cargo.toml Cargo.lock .