QA Engineer interview prep for Adobe.

Memory leaks in desktop or browser-based creative tools typically stem from detached DOM trees, lingering event listeners, unreleased WebGL textures, or retained references in long-lived undo/redo stacks. To detect leaks, create an automated stress scenario using Playwright or Puppeteer that repeatedly opens, edits, and closes documents across 50–100 iterations. At checkpoints, execute window.gc() (enabled with Node/V8 flags) and sample performance.memory.usedJSHeapSize or capture Chrome DevTools heap snapshots using CDP (Chrome DevTools Protocol). Analyze the heap snapshots programmatically to identify detached nodes and growing object retained sizes. If the baseline memory does not stabilize within an acceptable delta after garbage collection, flag a memory leak regression in CI before code merges.

API testing at Adobe requires strict contract adherence across multi-tenant enterprise customers. Implement consumer-driven contract testing using tools like Pact to verify that changes to microservices do not break existing downstream clients or SDKs. Complement contract tests with automated schema validation against OpenAPI or GraphQL schemas. For load and reliability, simulate peak traffic patterns using k6 or Locust to validate latency SLAs (p95 < 150ms), rate limiting, and graceful degradation during network partitions. In deployment pipelines, validate changes through canary releases and synthetic monitoring checks before rolling traffic out globally.

First, immediately reproduce the defect on a clean staging environment, document the exact reproduction steps, severity, and customer blast radius (e.g., potential data loss, crash rate, or affected browser configurations). Second, promptly raise the issue in the release escalation channel, notifying the engineering lead, product manager, and release manager with factual data rather than panic. Third, evaluate options collaboratively: determine if the offending feature can be turned off safely via an existing feature flag/kill switch without halting the release, whether a low-risk targeted patch can be cherry-picked, or if the release milestone must be postponed. Finally, after resolving the release, lead a blameless post-mortem to determine why automated CI checks did not catch the issue earlier and introduce regression tests to guarantee it cannot recur.