We currently have 27 benchmarks:

• 24 benchmarks can run up to 300 seconds each.

• 3 benchmarks run up to 10 seconds each.

In the worst-case scenario, where all benchmarks hit their upper time limits and some iterations spill over past the timeout (since we don't cancel in-progress iterations), the total runtime can reach:

• 24 × 300 = 7,200 sec

• 3 × 10 = 30 sec

• Total worst-case time: 7,230 sec (2h 30s)

While we never hit the worst-case, the recent addition of Client Bulk Write benchmarks has increased total execution time. In practice, we now exceed the 1-hour limit by 10–20 seconds, causing CI timeouts.

Acceptance criteria:

• Increase the benchmark timeout to 4000 seconds as a buffer to prevent CI failures.