time_delay resource is ignored when for_each loop is used in the calling code.

[jeffling-google]

Terraform CLI and Provider Versions

Terraform v1.4.5
on linux_amd64
hashicorp/google v4.63.0

Terraform Configuration

resource "google_access_context_manager_service_perimeter_resource" "service-perimeter-resource" {
  perimeter_name = local.vpc_sc_perimeter
  ...
}

resource "time_sleep" "wait_for_vpc_sc" {
  create_duration = "30s"
  depends_on      = [google_access_context_manager_service_perimeter_resource.service-perimeter-resource]
}

# Adding to Shared VPC as service projects
resource "google_compute_shared_vpc_service_project" "service_projects" {
  for_each        = { for project in local.projects : project.id => project }
  host_project    = each.value.shared_vpc_host
  service_project = each.value.id
  depends_on = [time_sleep.wait_for_vpc_sc]
}

Expected Behavior

The intention is to wait for 30 seconds and then start attaching projects to a Shared VPC host project.

Actual Behavior

The time_sleep resource does not take effect - TF tried to create resources right after the VPC-SC call. The time_sleep doesn't not even show up in terraform plan output. Once I removed the for_each statement so there is only one project is attached, the time_sleep resource shows up in the plan and works properly

Steps to Reproduce

Terraform plan will show you the differences between having for_each or no for_each.
Turn on TF_LOG=DEBUG, you will see that the time_sleep is not called.

How much impact is this issue causing?

High

Logs

No response

Additional Information

No response

Code of Conduct

• I agree to follow this project's Code of Conduct