Canonical claim

Collapse thresholds are primarily utilization-led.

The flagship study showed a shared transition band across five topology families. The controlled follow-up then showed that coupling mainly trims failure margin rather than relocating the threshold by much.

Across domains, collapse thresholds are primarily utilization-led; coupling mainly narrows the failure margin rather than moving the threshold much.

Flagship evidence

  • 5,000 synthetic systems across five topology families
  • Shared collapse transition band around baseline utilization 0.10 to 0.19
  • Median estimated threshold across families: 0.152
  • Variance growth and recovery lag generalized as early-warning signals
  • Heavy-tailed cascades, with lognormal fitting better than strict power law

Follow-up refinement

  • 2,700 fixed-size synthetic systems plus 144 healthcare variants
  • Mean high-versus-low coupling threshold shift: 0.008 utilization points
  • Largest absolute threshold shift: 0.023
  • Higher coupling reduced shock margin by about 39.2%
  • Healthcare validation showed the same cliff structure in arrival-multiplier space
Collapse probability vs utilization

Shared transition band

The strongest flagship result is the overlay itself: different network families cross the cliff at different exact points, but they still cluster inside the same narrow regime.

Phase transition by topology

Per-topology critical points

The topology differences are real, but they refine the claim rather than destroy it. The shared regime survives family-level variation.

Controlled thresholds

Controlled threshold sweep

When system size is fixed and utilization plus coupling are swept directly, the threshold remains utilization-led.

Threshold shift by coupling

Coupling moves the threshold only modestly

The follow-up weakens any stronger claim that coupling alone sets the universal threshold.

Failure shock margin

Coupling compresses robustness

The real operational effect of tighter coupling is not a dramatic threshold relocation. It is a narrower buffer to failure once a system is near the cliff.

Strongest defensible takeaway

The evidence supports a more careful claim than "one universal threshold rules everything." StressLab found a shared utilization-led collapse regime across domains. Coupling is important, but mostly because it makes near-threshold systems less forgiving.

That is useful both scientifically and operationally: the threshold tells you where the cliff is, while coupling tells you how little disturbance it may take to fall over it.