fix(whatsapp): retry webhook setup on transient failures before prompting reauthorization

app/jobs/channels/whatsapp/webhook_setup_job.rb

@@ -1,10 +1,21 @@
 class Channels::Whatsapp::WebhookSetupJob < ApplicationJob
   queue_as :low
 
-  # Runs Meta's phone-registration and webhook-subscription calls off the request
-  # thread. Inline, these Graph API calls can exceed the 15s Rack::Timeout and, since
-  # RequestTimeoutException bypasses setup_webhooks' rescue, abort inbox creation and
-  # roll it back — leaving the number connected on Meta but no inbox in Chatwoot.
+  # Meta's Graph API calls (phone registration + webhook subscription) are slow and fail
+  # transiently. Retry a few times, and once retries are exhausted mark the channel for
+  # reauthorization so the inbox has a visible recovery path instead of silently missing its
+  # webhook subscription. Running these off the request thread also keeps them clear of the
+  # 15s Rack::Timeout, which previously aborted inbox creation and rolled it back.
+  retry_on StandardError, wait: :polynomially_longer, attempts: 3 do |job, error|
+    channel = job.arguments.first
+    Rails.logger.error("[WHATSAPP] Webhook setup failed after retries: #{error.message}")
+    channel.prompt_reauthorization! if channel.is_a?(Channel::Whatsapp)
+  end
+
+  # A deleted channel can't be set up; discard instead of retrying (takes precedence over
+  # the StandardError retry above, which would otherwise catch DeserializationError too).
+  discard_on ActiveJob::DeserializationError
+
   def perform(whatsapp_channel, run_health_check: false)
     whatsapp_channel.setup_webhooks
     # Health check runs only after registration so a freshly provisioned number

app/models/channel/whatsapp.rb

@@ -120,13 +120,11 @@ class Channel::Whatsapp < ApplicationRecord
   delegate :media_url, to: :provider_service
   delegate :api_headers, to: :provider_service
 
-  # Runs synchronously inside Channels::Whatsapp::WebhookSetupJob; the Meta Graph
-  # calls are kept off the request thread so a slow response can't trip Rack::Timeout.
+  # Runs inside Channels::Whatsapp::WebhookSetupJob, off the request thread so the slow Meta
+  # Graph calls can't trip Rack::Timeout. Raises on failure so the job can retry the flaky
+  # calls and, once retries are exhausted, mark the channel for reauthorization.
   def setup_webhooks
     perform_webhook_setup
-  rescue StandardError => e
-    Rails.logger.error "[WHATSAPP] Webhook setup failed: #{e.message}"
-    prompt_reauthorization!
   end
 
   # Enqueue on the same channel record so GlobalID resolves it in the job. If the queue

spec/jobs/channels/whatsapp/webhook_setup_job_spec.rb

@@ -0,0 +1,49 @@
+require 'rails_helper'
+
+RSpec.describe Channels::Whatsapp::WebhookSetupJob do
+  let(:channel) do
+    create(:channel_whatsapp, validate_provider_config: false, sync_templates: false)
+  end
+
+  it 'runs webhook setup' do
+    allow(channel).to receive(:setup_webhooks)
+
+    described_class.perform_now(channel)
+
+    expect(channel).to have_received(:setup_webhooks)
+  end
+
+  it 'runs the provisioning health check when requested' do
+    allow(channel).to receive(:setup_webhooks)
+    allow(channel).to receive(:check_provisioning_health)
+
+    described_class.perform_now(channel, run_health_check: true)
+
+    expect(channel).to have_received(:check_provisioning_health)
+  end
+
+  it 'skips the provisioning health check by default' do
+    allow(channel).to receive(:setup_webhooks)
+    allow(channel).to receive(:check_provisioning_health)
+
+    described_class.perform_now(channel)
+
+    expect(channel).not_to have_received(:check_provisioning_health)
+  end
+
+  context 'when webhook setup keeps failing' do
+    before do
+      setup_service = instance_double(Whatsapp::WebhookSetupService)
+      allow(Whatsapp::WebhookSetupService).to receive(:new).and_return(setup_service)
+      allow(setup_service).to receive(:perform).and_raise(StandardError, 'meta down')
+    end
+
+    it 'retries and marks the channel for reauthorization once retries are exhausted' do
+      expect(channel.reauthorization_required?).to be false
+
+      perform_enqueued_jobs { described_class.perform_later(channel) }
+
+      expect(channel.reauthorization_required?).to be true
+    end
+  end
+end

spec/services/whatsapp/embedded_signup_service_spec.rb

@@ -67,18 +67,6 @@ describe Whatsapp::EmbeddedSignupService do
         expect(Rails.logger).to receive(:error).with('[WHATSAPP] Embedded signup failed: Token error')
         expect { service.perform }.to raise_error('Token error')
       end
-
-      # Webhook setup now runs in Channels::Whatsapp::WebhookSetupJob; the channel's own
-      # setup_webhooks rescue marks it for reauthorization when the Meta calls fail.
-      it 'marks the channel for reauthorization when webhook setup fails' do
-        real_channel = create(:channel_whatsapp, account: account, phone_number: '+1234567890',
-                                                 validate_provider_config: false, sync_templates: false)
-        allow(real_channel).to receive(:perform_webhook_setup).and_raise('Webhook setup error')
-
-        expect(real_channel.reauthorization_required?).to be false
-        real_channel.setup_webhooks
-        expect(real_channel.reauthorization_required?).to be true
-      end
     end
 
     context 'with reauthorization flow' do