docs(docs): List the shallower re-search after a root fail high as rejected

[aywrite]

Adds one entry to the "Measured and rejected" list in docs/ROADMAP.md.

The change searched a root fail high's re-search a ply shallower for each fail high since the depth opened or last failed low. It read -4 ±6 over 6,000 games at 10+0.1 against d82638a and failed sprt [0, 5] at LLR -3.11 on the last batch the cap allowed. The candidate was 53ee5ec, on claude/aspiration-window-research-1xdeqh, which stays pushed and unmerged.

Docs only; no engine code changes.

🤖 Generated with Claude Code

https://claude.ai/code/session_0133NvZMXKrttWcNCNU9xXjZ

---

Generated by Claude Code

[github-actions]

Speed against b48777af

Measured on AMD EPYC 7763 64-Core Processor.
Both sides built with rustc 1.99.0 (b940084d7 2026-09-28).

Both sides built and run on this runner in this job, the way
scripts/speed.sh measures a perf commit. Each round runs
both sides on a layout of its own, the same compiled code with
its code and data shuffled and moved, so the 95% interval carries
where the code landed as well as the run. The verdict holds the
interval against a 1% threshold. The default layout, the one a
release ships, is measured after as a diagnostic. The node
counts are the search's: they move when the search does, and a
speed change leaves them alone.

round     base nps  candidate nps  change
    1      4588744        4485474   -2.3%
    2      4587531        4533530   -1.2%
    3      4518603        4538100   +0.4%
    4      4602136        4527287   -1.6%
    5      4598421        4613193   +0.3%
    6      4464983        4532543   +1.5%
    7      4609230        4564184   -1.0%
    8      4499578        4527871   +0.6%
    9      4494745        4138174   -7.9%
   10      4595099        4132388  -10.1%
   11      4272810        4101813   -4.0%
   12      4131125        4033818   -2.4%
   13      4343505        4321946   -0.5%
   14      4555753        4448658   -2.4%
   15      4461889        4352849   -2.4%
   16      4545323        4507194   -0.8%
   17      4588702        4596029   +0.2%
   18      4532283        4608375   +1.7%
   19      4245234        4387523   +3.4%
   20      4576043        4403058   -3.8%
   21      4533682        4450744   -1.8%
   22      4517817        4485590   -0.7%
   23      4567077        4613069   +1.0%
   24      4448749        4566378   +2.6%
   25      4381440        4455387   +1.7%
   26      4456117        4369088   -2.0%
   27      4598054        4513940   -1.8%
   28      4542521        4481398   -1.3%
   29      4564286        4475117   -2.0%
   30      4527165        4498833   -0.6%
   31      4633602        4619857   -0.3%
   32      4534315        4479570   -1.2%
   33      4518712        4371668   -3.3%
   34      4558250        4405873   -3.3%
   35      4333983        4598821   +6.1%
   36      4559676        4624520   +1.4%
   37      4537738        4456518   -1.8%
   38      4553147        4443335   -2.4%
   39      4402349        4234576   -3.8%
   40      4488662        4459798   -0.6%

              nodes    time  median nps  faster half
base       11088536  2.45 s     4533998      4574782
candidate  11088536  2.47 s     4480484      4551809
change                            -1.2%        -0.5%

paired change -1.1%, 95% interval -1.8% to -0.4%
diagnostic, on the default layout alone -0.1%, 95% interval -1.9% to +1.7%, 13 rounds

not resolved: the interval reaches past ±1.0% without clearing it, so
more rounds are needed to say either way

Speed: -1.1% (bench nps, 95% interval -1.8% to -0.4%, 40 interleaved rounds over shuffled layouts vs b48777af)

The instructions each side's bench executed, counted under
cachegrind. The count repeats to within a few hundred
instructions, so a small change here is a real one, but it
prices instructions only: cache misses, mispredicted branches
and where the code lands are the speed's to show.

instructions against b48777af, one cachegrind run a side, bench

             instructions     nodes  per node
base       16,519,614,775  11088536    1489.8
candidate  16,519,614,947  11088536    1489.8
change             +0.00%              +0.00%

within ±0.7%, as far as edits not made for speed have moved the count