Test 1, change the medium, not the state: calibration (2026-10-14 12:29 Halifax).
Plan: plan.md, revision 2 (SHA-256 b2592cfea01be207022145da51d78b48c9f30dc1ff5dc53c0ae86026abe7f427).
Old starts 16 and 17 (seeds 15 and 16), ten moments each: 20 moments. These starts stay out of the pilot's results.
Coherence only, up to the restoration point R = T + 200, and whether the matched pushes and shams can be built. No gap growth is saved or looked at, and no restored arm is run after R.
C is the average of the two groups' coherence. The marks use the ratio of the averages over the 20 moments.
Marks: raising 1.5 to 3.0 times the untouched run's C at R; lowering at most 0.67 times; the push and its sham built at 18 or more of the 20 moments, raised and lowered separately. The smallest size meeting every mark is used unchanged in the pilot.

  size of K   raised: C(R) / untouched   lowered: C(R) / untouched   pushes built (raised, lowered)   marks met (raised, lowered, pushes)
      0.10%                      1.149                       0.884             20 and 20 of 20   no, no, yes
      0.25%                      1.390                       0.788             20 and 20 of 20   no, no, yes
      0.50%                      1.830                       0.630             20 and 20 of 20   yes, yes, yes
      1.00%                      3.382                       0.615             20 and 19 of 20   no, yes, yes
      2.00%                      5.445                       0.657             20 and 20 of 20   no, yes, yes
      5.00%                      8.517                       0.651             18 and 20 of 20   no, yes, yes
  chosen: 0.5% of K

At the chosen size (0.5% of K, so K = 1.9095 raised and 1.8905 lowered for 200 units):
  C(R) / untouched at single moments (shown only; the mark uses the ratio of averages): raised 0.86 to 3.68 (median 1.78); lowered 0.09 to 1.33 (median 0.69)
  matched pushes: amplitude (quarter turn = 1.571) raised median 0.0439, largest 0.1445; lowered median 0.0228, largest 0.0602
  shams: each group's coherence 0.9973 to 1.0031 times where it was
  built: raised 20 of 20, lowered 20 of 20
  average C during the window (20 moments), at T + 0, 50, 100, 150, 200:
    raised    0.0262  0.0315  0.0346  0.0433  0.0455
    lowered   0.0262  0.0217  0.0201  0.0196  0.0157
    untouched 0.0262  0.0259  0.0255  0.0281  0.0249

Kept-changed reference (the chosen size never restored, 700 units from T; coherence only). Reported here only; it isn't used in any pilot reading.
  average C in 100-unit blocks after T: (0, 100], (100, 200], (200, 300], (300, 400], (400, 500], (500, 600], (600, 700]
    raised    0.0311  0.0419  0.0527  0.0619  0.0649  0.0670  0.0686
    lowered   0.0229  0.0193  0.0172  0.0172  0.0164  0.0162  0.0164
    untouched 0.0262  0.0276  0.0252  0.0264  0.0273  0.0255  0.0259

Moments: start 16: T = 2045, 2375, 2705, 3035, 3365, 3695, 4025, 4355, 4685, 5015; start 17: T = 2195, 2525, 2855, 3185, 3515, 3845, 4175, 4505, 4835, 5165
Compute: 4 pieces, 8.6 minutes of foreground time (7 workers at a time). Machine: INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz.
The cause test's locked files unchanged: True.
Every calibration piece ran under this plan and this code: yes