Automated tests run by a TAS on a SUT can be subject to sudden bursts...

TAE highlights that logging must balance diagnostic value with execution performance and reliability. Direct synchronous file I/O for every log message can become a bottleneck during bursts, increasing latency and perturbing the timing of the automated interactions—especially for UI or time-sensitive integration tests—leading to flaky outcomes. Since all messages must be permanently stored, dropping burst logs (option C) violates the requirement. NTP synchronization (option A) helps correlate events across systems, but it does not address the performance overhead caused by bursty logging. The most useful approach is to buffer log events in memory and flush them periodically or asynchronously to disk. A circular buffer (or similar in-memory queue) reduces immediate I/O pressure and smooths bursts, while still preserving messages for later analysis when combined with an appropriate flush strategy and sizing. This design is aligned with TAE’s emphasis on making the TAS itself reliable and non-intrusive, ensuring logging supports triage without materially slowing or destabilizing test execution. Therefore, buffering in memory and periodically flushing to log files is the best solution.