sm121/launcher @ e8ac5f1bd727ecbd3d0ecda374e518c7d49be8f4

Alexander Korolev d1f10dd1e7 bench: r28 full rerun 159/176 — 155 vs 160 was run-to-run variance (8 flips), not a regression		преди 1 месец
..
reports	d1f10dd1e7 bench: r28 full rerun 159/176 — 155 vs 160 was run-to-run variance (8 flips), not a regression	преди 1 месец
recipe.toml	ff1f727ae3 recipes: GMU 0.88 in r27/r28 4x+2x (startup gate 107.0 vs 109.5 GiB; KV pinned, capacity unchanged)	преди 1 месец