ChromeDriver 127 closes beforeunload prompts immediately in HTTP sessions; BiDi mode with unhandled_prompt_behavior: { default: 'ignore' } keeps them open for automated tests.

Use a local gem copy in Gemfile to test code changes immediately and debug or modify the gem without publishing. Avoid committing local paths because they break for other developers.

Rails jobs can be hard to inspect because adapters behave differently in development, test, and production. Logging, synchronous execution, and queue tables help trace failures and pending work.

aggregate_failures in RSpec groups multiple expectation failures in one example, improving test feedback and making debugging mismatched values easier.

ActiveJob concurrency checks can pass falsely with the :test adapter because blocked jobs still enqueue; using :solid_queue verifies real blocking behavior.

Time comparisons can fail in tests because Ruby keeps subsecond fractions even when the printed values look identical.

CarrierWave defaults often need project-specific tweaks for storage paths, cache cleanup, test isolation, and large-file handling.

Rails 7.2 changes show_exceptions defaults in tests, so rescued RecordNotFound often becomes a 404 response instead of a raised exception.

Browser and driver mismatches can break Selenium tests; Selenium Manager can fetch matching Chrome and chromedriver versions automatically, though the first run is slower.

Code coverage reveals untested Rails paths and blind spots across RSpec, Cucumber, and parallel test runs. SimpleCov can separate reports and add branch coverage.

Randomized RSpec order can expose order-dependent failures that pass alone but fail in full runs; rspec --bisect finds a minimal reproducer for the offending interaction.

Async rejections can fail Jasmine tests as unhandled promise rejections when the promise is not awaited or returned. jasmine.spyOnGlobalErrorsAsync() lets tests ignore or inspect the global error.

Long RSpec examples are hard to scan and slow to report progress. A small STEP helper adds semantic section markers and clearer test output.

Avoid duplicated Cucumber scenarios by sharing setup and assertions with step composition, helper modules, or scenario outlines. Better reuse keeps integration tests DRY and stack traces readable.

External libraries add maintenance and security risk, so picking a Ruby gem needs attention to necessity, test quality, license, maturity, and scope fit.

Exact fractions in Ruby prevent rounding drift in unit conversions and tests, but mixing Rational with BigDecimal or floats can still reintroduce inaccuracies.

Flaky RSpec examples can be retried a limited number of times in CI instead of failing immediately, using rspec-rebound with configurable retry settings.

Rerun failing Cucumber scenarios automatically or from tmp/parallel_cucumber_failures.log when a test run leaves broken features behind.

Stricter strong parameters raise missing and unpermitted attribute errors during development and tests, helping catch bad requests earlier while requiring environment-specific handling in production.

Business-rule tests with multiple variables can explode into 2n permutations. An n+1 pattern keeps coverage manageable while making individual cases easier to follow.

Strict loading in index actions catches N+1 queries when views touch unpreloaded associations, causing test failures instead of hidden database hits.

Repeated RSpec assertions can be extracted into custom matchers, but some cases are better served by private helper methods when reuse would be too awkward.

Random behavior in Jasmine tests makes results unstable unless spyOn controls dependent calls. returnValue, returnValues, and callFake keep shuffles predictable.

Remote Chrome debugging in Docker is awkward when integration tests need a browser you can inspect. SSH port forwarding can expose a local chromedriver inside the test container.