[gvegayon]

Pins 07ed7a0d6243653c5c197cb8c7420baad6d279c5 (UofUEpiBio/epiworld#353).

3d-3b moves Bubbles onto a household layer and adds an example model. These scenarios don't use Bubbles, so the epiworld-v1 runners are unchanged and the code they run is the same as at 3d-3a. The run used the same image (9a51e13558f5), both C++ engines, with replicates interleaved. make container-check and make container-smoke pass, and all 1,280 runs are ok.

Rows

Each cell is 0.x, then 1.0. Simulate and build are in seconds, peak in MiB.

Scenario	n	Simulate	Ratio	Build	Ratio	Peak	Ratio	Attack rate
00	10,000	0.00483, 0.00486	1.01	0.00134, 0.00109	0.82	8, 8	0.95	0.385, 0.385
00	100,000	0.00917, 0.00947	1.03	0.0139, 0.0107	0.77	38, 30	0.79	0.060, 0.060
01	10,000	0.00242, 0.00243	1.00	0.00132, 0.00108	0.82	8, 8	0.94	0.119, 0.119
01	100,000	0.00982, 0.00957	0.97	0.014, 0.0109	0.78	43, 38	0.88	0.014, 0.014
02	10,000	0.00242, 0.00243	1.00	0.00131, 0.00108	0.82	9, 8	0.95	0.119, 0.119
02	100,000	0.01, 0.00964	0.96	0.0142, 0.0112	0.79	44, 38	0.88	0.014, 0.014
03	1,000,000	0.033, 0.0301	0.91	0.167, 0.129	0.77	352, 253	0.72	0.006, 0.006
04	165,865	0.24, 0.202	0.84	0.0192, 0.0136	0.71	90, 84	0.93	0.622, 0.622

Ratios versus 3d-3a

One ratio grew by more than 10%: the simulation ratio of scenario 03 (1,000,000 agents), from 0.81 to 0.91. It is not a 1.0 regression:

• The 1.0 time fell, from 0.0317 s to 0.0301 s.
• The ratio rose because the 0.x engine's time fell, from 0.0391 s to 0.033 s, and its code did not change.
• This row has been noisy before: it was 0.94 at 3d-1b.
• The epiworld code these runners use is identical between 3d-3a and 3d-3b.

So I did not profile it. Every other ratio is within 10% of 3d-3a; scenario 04's simulation ratio improved from 0.91 to 0.84. Peak memory is unchanged, and attack rates agree in every row.

The README and docs figures were not re-rendered (the 3d-3a PR re-rendered them); this PR has the records, the history, the table and the environment record only.

🤖 Generated with Claude Code