Problem

At BuildForge, a test is considered flaky if it has both passed and failed across multiple CI runs. Given a list of test execution records, return the names of all flaky tests in lexicographical order.

Implement a function that scans the records and detects which tests are inconsistent over time.

Formal Specification

• Input: records, a list of pairs [test_name, status]
  • test_name is a non-empty string
  • status is either "pass" or "fail"
• Output: A list of unique test names that have appeared with both statuses at least once

Examples

Example 1

Input: records = [["login", "pass"], ["checkout", "fail"], ["login", "fail"], ["search", "pass"]]
Output: ["login"]

login appears once as pass and once as fail, so it is flaky.

Example 2

Input: records = [["a", "pass"], ["b", "fail"], ["a", "pass"], ["b", "fail"]]
Output: []

Neither test changes outcome, so no test is flaky.

Constraints

• 1 <= len(records) <= 10^5
• 1 <= len(test_name) <= 100
• Each status is either "pass" or "fail"
• Test names may appear multiple times
• Output must contain unique names sorted in ascending lexicographical order