Docs changes which essentially suggest not to use this feature.

third_party/terraform/utils/provider.go.erb

@@ -103,7 +103,7 @@ func Provider() terraform.ResourceProvider {
 				Optional: true,
 			},
 
-			"synchronous_timeout": {
+			"request_timeout": {
 			    Type:     schema.TypeString,
 			    Optional: true,
 			},
@@ -412,7 +412,7 @@ func providerConfigure(d *schema.ResourceData, terraformVersion string) (interfa
 		terraformVersion:    terraformVersion,
 	}
 
-	if v, ok := d.GetOk("synchronous_timeout"); ok {
+	if v, ok := d.GetOk("request_timeout"); ok {
 		var err error
 		config.RequestTimeout, err = time.ParseDuration(v.(string))
 		if err != nil {

third_party/terraform/website/docs/guides/provider_reference.html.markdown

@@ -112,6 +112,11 @@ Values are expected to include the version of the service, such as
 * `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.
 
+* `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
+amount of time the provider will wait for a logical operation - use the resource
+timeout blocks for that.
+
 The `batching` fields supports:
 
 * `send_after` - (Optional) A duration string representing the amount of time
@@ -331,6 +336,18 @@ Valid time units are "ns", "us" (or "µs"), "ms", "s", "m", "h".
 * `disable_batching` - (Optional) Defaults to false. If true, disables global
 batching and each request is sent normally.
 
+---
+* `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
+amount of time the provider will wait for a logical operation - use the resource
+timeout blocks for that.  This will adjust only the amount of time that a single
+synchronous request will wait for a response.  The default is 30 seconds, and
+that should be a suitable value in most cases.  Many GCP APIs will cancel a
+request if no response is forthcoming within 30 seconds in any event.  In
+limited cases, such as DNS record set creation, there is a synchronous request
+to create the resource.  This may help in those cases.
+
+
 ---
 
 * `user_project_override` - (Optional) Defaults to false. If true, uses the