These timings were collected locally on Windows/Python 3.13.15 from clean commit 6373d23. The earlier dirty Python 3.14 measurements are not reused here.

The hosted manual workflow attempt at this same revision completed its scientific scorecard but could not start ASV timings because the fresh runner lacked a local main branch. That setup failure is retained in hosted-workflow.log; it is distinct from the local timing results below. The Linux results page uses a later revision with the ASV default changed to HEAD. Do not pool these environments or interpret differences as an OS performance comparison.

The local regression log records 21 tests run, with one documentation-build check skipped because no HTML had yet been built in the clean checkout. Hosted correctness and documentation checks for this exact revision passed separately.

This Windows capture records a platform CPU label and logical CPU count, but its custom ASV machine label omitted the full CPU/RAM inventory. Those missing fields remain explicit. The corrected hosted workflow captures the complete automatic machine inventory.
