Check 2 of the pilot review: do moments with bigger slower-half or timing differences show bigger medium-versus-push growth differences? (run 2026-10-14 15:05 Halifax)
After the fact and exploratory: a description of the pilot's saved results, not a reading fixed in plan.md and not a new experiment. No model was run.
A toy model: structure inside the model, not evidence about cells. Histories, not memory: differences are between intervention histories and the states they leave.
Read: the pilot's 120 moment files (starts 103 to 114), each matching the hash in pilot_results.json; 119 moments with both pushes built. Reproduced before reading on: growth difference (S difference) (largest difference 1.1e-19); size of the coherence change (dC) (largest difference 0.0e+00); slower-half and timing descriptions at R (largest difference 0.0e+00); timing difference at R (check 1) (largest difference 5.6e-17).
Each difference is medium minus matched push, in the form of the report's S difference: positive is the direction seen on average. Correlations are within each start, averaged over the 12 starts. Error bars and 'clear' (more than 3 bars) as in the report.

Answer: slower half: not clear; timing: not clear.

Averages (like for like)
  growth difference Y: +0.00181 +- 0.00051 (+3.6 bars; 10 of 12 starts above zero)
    raised arm: +0.00306 +- 0.00067 (+4.6 bars; 12 of 12 starts above zero); lowered arm: +0.00055 +- 0.00062 (+0.9 bars; 8 of 12 starts above zero)
  slower-half difference X_slow: +0.0170 +- 0.0008 (+22.3 bars; 12 of 12 starts above zero)
  timing difference X_time (radians): +0.4071 +- 0.0350 (+11.6 bars; 12 of 12 starts above zero)
  size of the coherence change dC: +0.0145 +- 0.0006 (+26.0 bars; 12 of 12 starts above zero)

Slower half (X_slow)
  correlation with the growth difference (the main measure): +0.064 +- 0.104 (+0.6 bars; 8 of 12 starts above zero)
    with the size of the coherence change held fixed: -0.085 +- 0.095 (-0.9 bars; 5 of 12 starts above zero)
    raised arm alone: -0.094 +- 0.099 (-0.9 bars; 5 of 12 starts above zero)
    lowered arm alone: +0.193 +- 0.096 (+2.0 bars; 9 of 12 starts above zero)
  growth difference, biggest-difference half of moments minus smallest: -0.00039 +- 0.00104 (-0.4 bars; 6 of 12 starts above zero)
  share of the average growth difference the link would account for: +0.31 +- 0.59 (pooled within-start slope +0.0328; an extrapolation, meaningful only if the link is clear)

Timing (X_time)
  correlation with the growth difference (the main measure): +0.168 +- 0.083 (+2.0 bars; 7 of 12 starts above zero)
    with the size of the coherence change held fixed: +0.119 +- 0.089 (+1.3 bars; 8 of 12 starts above zero)
    raised arm alone: +0.222 +- 0.076 (+2.9 bars; 8 of 12 starts above zero)
    lowered arm alone: +0.062 +- 0.115 (+0.5 bars; 6 of 12 starts above zero)
  growth difference, biggest-difference half of moments minus smallest: +0.00033 +- 0.00093 (+0.3 bars; 6 of 12 starts above zero)
  share of the average growth difference the link would account for: +0.35 +- 0.20 (pooled within-start slope +0.001554; an extrapolation, meaningful only if the link is clear)

How entangled the measures are (correlations within starts)
  slower half with timing: +0.337 +- 0.074 (+4.5 bars; 10 of 12 starts above zero)
  slower half with the size of the coherence change: +0.832 +- 0.034 (+24.5 bars; 12 of 12 starts above zero)
  timing with the size of the coherence change: +0.327 +- 0.048 (+6.8 bars; 11 of 12 starts above zero)
  growth difference with the size of the coherence change: +0.136 +- 0.119 (+1.1 bars; 7 of 12 starts above zero)

Limits
  Each moment's growth difference is noisy, so only fairly strong links can be clear here; 'not clear' does not show that a feature doesn't matter.
  A link across moments is not cause: moments where the medium acted more strongly could show bigger differences of every kind. Holding the size of the coherence change fixed addresses only that one.
  The slower-half and timing differences are read at R only; the individual oscillators after R weren't saved.

Compute: 21.4 seconds, in the foreground.