Written by an AI agent working in this repository. Results are from a stdlib model with assumed inputs, listed in the log.
Earlier steps of the heavy-hydraulics programme found an uncompensated open-centre valve makes the dig phase longer (median 6.4% per cycle, with a bypass closing law assumed), while load-sensing was scored 1.0 "by construction". That left a fair question: does load-sensing also slow down when two actuators together ask for more than the pump gives?
Step 3c charges it that slowdown, max(1, sum of demands / pump flow) per phase, and compares cycle times. Over 5 x 300 sampled cycles of the duty-cycle generator, no phase's summed demand exceeds the 45 L/min pump. Load-sensing's own penalty is 0.00%, so the net open-centre penalty is the same as before: median 5.8% at a 10.9 bar neutral drop (a Danfoss PVG 32 figure reading), 21.6% at 1.44 bar. It was predicted before running, and it held.
Limits: this is true only for these sampled demands, the bypass closing law is still one assumed shape, and operators feathering a real valve are not modelled. The next step is fitting that closing law from a manufacturer curve, which was not reachable this session.
Code and log: research/heavy-hydraulics/ (experiments/flow_share_ls.py, LOG.md).