Reversing contact order within 20-second bins changes 3.7–22.6% of reachable pairs in five-minute sequential spread tests on six human contact networks

We test whether contact order within a 20-second bin changes modeled reach. We use 3,951 complete, nonempty five-minute windows from six SocioPatterns contact networks. Spread is deterministic, allows waiting, and has no recovery. Two sequential rules immediately update endpoints after each contact. One sorts contacts by anonymous endpoint IDs. The other reverses that order within each bin. Total reachable-pair counts differ by 0.085% to 0.532% across datasets. Yet 3.7% to 22.6% of ordered pairs in the union occur under only one order. Each pair is counted separately in each window. A second comparison permits one step per bin or all steps within each snapshot component. The latter adds 4.3% to 50.5% more reachable pairs. A shifted five-minute grid gives similar gaps. At 60 minutes, relative gaps range from 0.7% to 17.2%. Independent single-source code agrees on 1,000 random graphs and 90 real windows. The strict and non-strict path definitions are established prior work. This contribution supplies an empirical comparison and code to check update rules. It does not identify real within-bin timing or estimate disease risk. The package contains source code, aggregate tables, figures, source hashes, and reproduction instructions. The source archive has a non-commercial licence and must be retrieved separately. The union-normalized pair disagreement was added after the first results as a descriptive measure.

temporal networks · human contact · reachability · simultaneous contacts · sequential updates · reproducibility · SocioPatterns