Shipping products with confidence: how we rebuilt our testing tools for velocity
30 Sep 2026, 8:45 am (UTC)
45 Minutes
Speakers
A Drupal-based product’s test suite can consume six days of GitHub runner time and still include flaky tests across more than 500 jobs. After a decade of accumulated complexity, Open Social revisits its testing foundation, using the testing pyramid to evaluate what the suite tests, at what cost, and how much confidence those tests provide.
The rebuilt tooling combines a hidden, environment-gated HTTP Test Control Interface (TCI) in Drupal with Playwright as the test harness. Drupal modules declare attribute-driven PHP QA operations, which the TCI publishes as an OpenAPI specification, separating test data setup from the runner. Playwright provides trace files, automatic retries, and parallel execution; the session also covers how its testing insights help debug failures.
Cucumber provides human-readable UI test specifications, while a Markdown-based format supports GraphQL tests with embedded queries, variables, and expectations. The session presents an architecture for rebuilding testing confidence incrementally, without replacing the entire stack at once.