From 6b63a67a0c4b074f7fc826225a5a2b8bd781e805 Mon Sep 17 00:00:00 2001
From: Anso <dev@saelix.com>
Date: Fri, 19 Jun 2026 21:45:04 -0400
Subject: [PATCH] ci: retry E2E tests in CI to absorb flaky real-Docker timing
 (#1396)

The E2E suite includes real-Docker deploy and health-gate specs whose
timing varies on shared CI runners (cold image pulls, gate observation
windows). With no retries configured, a single transient timing blip
failed the whole Playwright job. Enable 2 retries in CI only; local runs
stay at 0 for fast, deterministic feedback. The existing
trace: 'on-first-retry' now captures the retried attempt for triage.
---
 playwright.config.ts | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/playwright.config.ts b/playwright.config.ts
index 404983c5..6faa7913 100644
--- a/playwright.config.ts
+++ b/playwright.config.ts
@@ -19,6 +19,10 @@ export default defineConfig({
   expect: { timeout: 5_000 },
   // Run tests serially - Sencho is a single-user app and tests share DB state
   workers: 1,
+  // Retry only in CI: the deploy/health-gate specs drive real Docker container
+  // runs, so image-pull and gate-window timing varies on shared runners. A retry
+  // absorbs the transient blip; trace: 'on-first-retry' captures the retried run.
+  retries: process.env.CI ? 2 : 0,
   reporter: [['list'], ['html', { outputFolder: 'e2e/report', open: 'never' }]],
 
   use: {