Given the explicit risk that the web app may crash during execution, the highest-priority tool capability is resilience: the ability to recover, continue, and provide usable results from unattended nightly runs. TAE emphasizes that automation must be reliable as a process, not just at the single-test level. If one crash aborts the entire suite, the organization loses feedback for many tests, reduces confidence in the pipeline, and increases triage cost. Therefore, capabilities such as automatic restart of the browser/app, test isolation, robust teardown, failure handling, skipping/marking affected tests, and resuming execution with proper reporting are critical evaluation criteria. Option A (descriptive meta-language) can help readability or non-coder authoring but is not the most urgent need based on the scenario. Option C (mock server) is useful for isolating dependencies in some test levels, but the scenario is UI tests against the most recent build; nothing indicates an API dependency problem that drives tool selection here. Option D (licensing feature sets) affects procurement, but it does not directly mitigate the stated operational risk. Hence, recovery and continuation support is the most important concern.