Walk & Test: add 0.25s and 0.5s round intervals

[TechLuddite]

Adds 0.25s and 0.5s to the round-interval dropdown. Both were measured before being offered rather than assumed to work.

The browser sustains it

At a 250 ms gap the loop settles at 0.306 s per round, and the number does not move as the walk deepens:

Window	Rounds	Per round	Request rate
0-15s	49	0.306 s	32.7 req/s
15-30s	49	0.306 s	32.7 req/s
30-60s	97	0.309 s	32.3 req/s
60-90s	97	0.309 s	32.3 req/s
90-120s	97	0.309 s	32.3 req/s

389 rounds, 3,890 samples. Sort stayed responsive at 159 ms and opening a new spot at 280 ms with that many samples behind them. The saved record came to 95 KB after truncation.

The concern going in was that per-round aggregation rescans every sample, so cost would grow with walk length. At these depths it does not, and the measurement says so rather than an optimisation nobody needed.

It holds because the loop awaits a whole round before starting the interval timer, so the setting is a gap between rounds rather than a fixed cadence. Requests cannot pile up, and a slow link stretches the gap instead.

What it is not is polite

Sustained, a quarter-second gap is up to about 2,000 requests to each of ten third parties over ten minutes. Selecting either sub-second option now raises a notice saying so, which also names the failure mode that matters for reading the result: a destination whose bot protection starts refusing appears as a dead card when it is really a rate limit.

SUB_SECOND_INTERVAL_MS carries the threshold and the reasoning, and a test pins exactly which choices it flags — safe for the browser and polite to other people's servers are separate questions, and the notice is about the second one.

Also

The disclosure list and the README privacy table carry the revised request volume; the previous figures predated anything faster than one second.

Verification

• npm run check green: 314 tests, 1 new.
• Offline regression re-run: still zero occurrences of "N ms" anywhere on the page.
• Notice confirmed rendering at 0.25s in Chromium; estimate rounds to the nearest hundred and errs high, since understating a cost you are warning about is the wrong direction.

---

Generated by Claude Code