Pause test: start 7 with the nudge test's run code, in one go (A) and with one pause (B)
Run 2026-10-13 11:35 to 11:37 Halifax (A 150 s, B 149 s). B paused at t = 180 and picked up again from its kept progress.

A with B: settling same; first gap difference at t = 185; first coherence difference at t = 1,720.
    t = 0 to 1,000: largest gap log difference 1.40e-08; largest coherence difference 0.00e+00 (typical coherence 0.0291; r = 1.000)
    t = 1,000 to 2,000: largest gap log difference 9.52e-08; largest coherence difference 1.82e-09 (typical coherence 0.0172; r = 1.000)
    t = 2,000 to 3,000: largest gap log difference 6.49e-04; largest coherence difference 2.47e-05 (typical coherence 0.0300; r = 1.000)
    t = 3,000 to 4,000: largest gap log difference 3.69e-01; largest coherence difference 7.16e-02 (typical coherence 0.0228; r = 0.673)
    t = 4,000 to 5,000: largest gap log difference 2.64e-01; largest coherence difference 6.24e-02 (typical coherence 0.0205; r = -0.027)
    t = 5,000 to 6,000: largest gap log difference 2.96e-01; largest coherence difference 7.79e-02 (typical coherence 0.0320; r = -0.329)
A with the saved run: settling same; first gap difference at t = 320; first coherence difference at t = 1,834.
    t = 0 to 1,000: largest gap log difference 2.80e-08; largest coherence difference 0.00e+00 (typical coherence 0.0291; r = 1.000)
    t = 1,000 to 2,000: largest gap log difference 2.18e-07; largest coherence difference 1.37e-09 (typical coherence 0.0172; r = 1.000)
    t = 2,000 to 3,000: largest gap log difference 1.30e-04; largest coherence difference 3.80e-06 (typical coherence 0.0300; r = 1.000)
    t = 3,000 to 4,000: largest gap log difference 1.75e-01; largest coherence difference 2.85e-02 (typical coherence 0.0228; r = 0.900)
    t = 4,000 to 5,000: largest gap log difference 3.14e-01; largest coherence difference 6.41e-02 (typical coherence 0.0205; r = 0.060)
    t = 5,000 to 6,000: largest gap log difference 2.01e-01; largest coherence difference 6.36e-02 (typical coherence 0.0320; r = 0.105)
B with the saved run: settling same; first gap difference at t = 185; first coherence difference at t = 1,720.
    t = 0 to 1,000: largest gap log difference 1.90e-08; largest coherence difference 0.00e+00 (typical coherence 0.0291; r = 1.000)
    t = 1,000 to 2,000: largest gap log difference 1.35e-07; largest coherence difference 1.82e-09 (typical coherence 0.0172; r = 1.000)
    t = 2,000 to 3,000: largest gap log difference 5.24e-04; largest coherence difference 2.28e-05 (typical coherence 0.0300; r = 1.000)
    t = 3,000 to 4,000: largest gap log difference 3.77e-01; largest coherence difference 6.87e-02 (typical coherence 0.0272; r = 0.665)
    t = 4,000 to 5,000: largest gap log difference 3.07e-01; largest coherence difference 5.71e-02 (typical coherence 0.0267; r = 0.012)
    t = 5,000 to 6,000: largest gap log difference 1.67e-01; largest coherence difference 5.94e-02 (typical coherence 0.0241; r = -0.092)

Reading, as set before the run: B parts from A, so on this machine picking up again changes the numbers by itself. The pauses are enough to explain the paused batches' departures; that their pauses caused them is not shown.
Layout: built uf strides (4096, 16, 32); read back in C order (4096, 2048, 16).
Folders and locked files unchanged: lock yes, cause_explore yes, nudge yes, growth yes