diff --git a/envs/bootstrap/outputs.tf b/envs/bootstrap/outputs.tf
index 6d30484..c7a771d 100644
--- a/envs/bootstrap/outputs.tf
+++ b/envs/bootstrap/outputs.tf
@@ -4,8 +4,8 @@ output "state_bucket" {
 }
 
 output "init_command" {
-  description = "Ready-made init for the environment, since a backend block cannot take variables."
-  value       = "tofu -chdir=envs/lance.blue init -backend-config=\"bucket=${aws_s3_bucket.state.id}\" -backend-config=\"region=${var.region}\""
+  description = "Ready-made init for the environment. envs/lance.blue/backend.tf writes the bucket and region out in full, since a backend block cannot take variables, so this needs no -backend-config flags."
+  value       = "tofu -chdir=envs/lance.blue init"
 }
 
 output "apply_role_arn" {
diff --git a/modules/match-cluster/variables.tf b/modules/match-cluster/variables.tf
index 649b85e..9cd1800 100644
--- a/modules/match-cluster/variables.tf
+++ b/modules/match-cluster/variables.tf
@@ -40,7 +40,7 @@ variable "task_cpu" {
 }
 
 variable "task_memory" {
-  description = "Fargate memory in MiB. Must be a combination Fargate allows for the chosen CPU. The spike measured ~730MB RSS per match, but that was the headless server and bots - the Suramadu server and the client JVM are on top of it and have never been measured together, so 2048 is a guess."
+  description = "Fargate memory in MiB. Must be a combination Fargate allows for the chosen CPU. The spike measured ~730MB RSS per match, but that was only the headless server and bots. Suramadu's server and the client JVM have since been measured together: a full match peaked at 3.2GB (see envs/lance.blue/variables.tf's match_task_memory and docs/architecture.md), and lance.blue's env now sets 8192 to clear arena's heap ceilings. 2048 remains this module's generic default."
   type        = number
   default     = 2048
 }