Particle (Lagrangian) stress history differs from serial by 1.2e-5 at np >= 4

[lmoresi]

Symptom

On the turned-over Maxwell box of tests/parallel/test_1062 (shear modulus varying in x, two counter-rotating cells between no-slip walls, unstructured simplex mesh, 10 steps), the shear stress at (0.5, 0.2) with stress_transport="lagrangian" is

ranks	sxy
1	-0.1214269
3	-0.1214269
4	-0.1214385
6	-0.1214385

Every other stress history (backward/forward from nodes and integration points, the SUPG grid) gives the serial value to 1e-7 at np 3, 4 and 6 on feature/forward-from-nodes (b20c992 and after); tests/parallel/test_1066 holds them to 1e-6 and marks this case as an expected failure at np >= 4.

What is ruled out

• the flux evaluated at the particles: no point extrapolated, every particle owned by the rank that evaluates it, the same 3220 particles in total;
• the cell proxy's fit: no thin or empty cells (at least 5 particles per cell);
• point location: every departure point has exactly one owner.

The stress read at step 0 already differs by 1e-8, and the velocity at step 1 by 7e-6.

Reproduction

mpirun -n 4 python -m pytest --with-mpi tests/parallel/test_1066_transport_schemes_mpi.py -k lagrangian

Underworld development team with AI support from Claude Code