Benchmark bug hunt reveals Python's 4,300-digit string limit

A developer discovered their A2A benchmark stopped at N=22 because CPython 3.11's default int-to-string conversion cap (4,300 digits) crashed the Python agent on the 24th Mersenne prime (6,002 digits). Removing a decorative str() call restored the missing data point. The investigation also uncovered fragile regex parsing of LLM output and zero-value timing bugs in Rust.