diff --git a/third_party/terraform/utils/batcher.go b/third_party/terraform/utils/batcher.go
index caa6b45958ab..9f1a0ecdcb9e 100644
--- a/third_party/terraform/utils/batcher.go
+++ b/third_party/terraform/utils/batcher.go
@@ -10,7 +10,7 @@ import (
 	"github.com/hashicorp/errwrap"
 )
 
-const defaultBatchSendIntervalSec = 10
+const defaultBatchSendIntervalSec = 3
 
 // RequestBatcher is a global batcher object that keeps track of
 // existing batches.
diff --git a/third_party/terraform/website/docs/guides/provider_reference.html.markdown b/third_party/terraform/website/docs/guides/provider_reference.html.markdown
index 35ef8546b2e2..b6de0b0b4e2b 100644
--- a/third_party/terraform/website/docs/guides/provider_reference.html.markdown
+++ b/third_party/terraform/website/docs/guides/provider_reference.html.markdown
@@ -110,7 +110,8 @@ Values are expected to include the version of the service, such as
 `https://www.googleapis.com/compute/v1/`.
 
 * `batching` - (Optional) This block controls batching GCP calls for groups of specific resource types. Structure is documented below.
-~>**NOTE**: Batching is not implemented for the majority or resources/request types and is bounded by the core [`-parallelism`](https://www.terraform.io/docs/commands/apply.html#parallelism-n) flag. Adding or changing this config likely won't affect a Terraform run at all unless the user is creating enough of a particular type of resource to run into quota issues.
+~>**NOTE**: Batching is not implemented for the majority or resources/request types and is bounded by two values. If you are running into issues with slow batches
+resources, you may need to adjust one or both of 1) the core [`-parallelism`](https://www.terraform.io/docs/commands/apply.html#parallelism-n) flag, which controls how many concurrent resources are being operated on and 2) `send_after`, the time interval after which a batch is sent.
 
 * `request_timeout` - (Optional) A duration string controlling the amount of time
 the provider should wait for a single HTTP request.  This will not adjust the
@@ -120,7 +121,8 @@ timeout blocks for that.
 The `batching` fields supports:
 
 * `send_after` - (Optional) A duration string representing the amount of time
-after which a request should be sent. Defaults to 10s.
+after which a request should be sent. Defaults to 3s. Note that if you increase
+`parallelism` you should also increase this value.
 
 * `enable_batching` - (Optional) Defaults to true. If false, disables batching
    so requests that have batching capabilities are instead is sent one by one.