Issues

• On the closed-form Jacobian a plain run takes 10x the steps where the GH #135 clamp fires: the retry keeps J's slope in a slot it reads at 0 (MODEL2001200002: 5,011 steps, 478 on the difference quotient)

  performance
  lanl/bngsim#1466 · Filter by author wshlavacek opened on Nov 2, 2026

• SSA observables and expressions read count/V_static in a resized compartment: B := 2*X reports 64 for 32 after c: 2 -> 4

  bugcorrectness concernfamily: variable volumes
  lanl/bngsim#1464 · Filter by author wshlavacek opened on Nov 2, 2026

• SSA starts from the load-time count when an event-resized compartment's starting size is written: X = 100 molecules for 150

  bugcorrectness concernfamily: variable volumes
  lanl/bngsim#1463 · Filter by author wshlavacek opened on Nov 2, 2026

• Fit gradients: four small cleanups left over from #1376

  cleanuplow priority
  lanl/bngsim#1460 · Filter by author wshlavacek opened on Nov 2, 2026

• differential: on klu_vs_dense and cc_vs_interp the third arms rerun the two sides, so a bug on one side reads split or is blamed on the other

  questiontesting
  lanl/bngsim#1459 · Filter by author wshlavacek opened on Nov 2, 2026

• Leftovers from #1380: a second copy of the rate-law walk, and _lambdify_doubles cannot print exp

  lanl/bngsim#1458 · Filter by author wshlavacek opened on Nov 2, 2026

• A root CVODE places inside a step that went past a bend's corner restarts the columns from the step's interpolant: 1e-4 to 1e-2 off on some output grids

  bugfamily: sensitivity terms
  lanl/bngsim#1456 · Filter by author wshlavacek opened on Nov 2, 2026

• The clock-rate column of a narrow pulse whose condition reads time() does not converge: 3.8e-4 at rtol 1e-12, its corners carried rather than rooted

  bugfamily: sensitivity terms
  lanl/bngsim#1455 · Filter by author wshlavacek opened on Nov 2, 2026

• A narrow pulse with a jump leaves the clock-rate column δt behind the state at each edge: 4.6e-5 at w = 0.003, at every tolerance

  bugfamily: sensitivity terms
  lanl/bngsim#1454 · Filter by author wshlavacek opened on Nov 2, 2026

• The #926 slide detector reads its band as the restart test did before #1329 and #1450: static atol, species one by one, 10·√N

  low priority
  lanl/bngsim#1452 · Filter by author wshlavacek opened on Nov 2, 2026

• GH #176's parked fixture and its twins land on the wrong branch, about 63% off with no error, at rtol 3e-8 and looser

  correctness concernquestion
  lanl/bngsim#1451 · Filter by author wshlavacek opened on Nov 2, 2026

• nfsim: integrating a linear time table's ramp refreshes every propensity, even ones no table reaches

  low priorityperformance
  lanl/bngsim#1449 · Filter by author wshlavacek opened on Nov 2, 2026

• jax: on a difference-quotient model, differentiable_solve's value and its gradient come from runs at different tolerances

  lanl/bngsim#1446 · Filter by author wshlavacek opened on Nov 1, 2026

• A state that leaves a rate-law equality at once after a clock-switch stop is refused, where the columns are exact

  bugfamily: sensitivity terms
  lanl/bngsim#1445 · Filter by author wshlavacek opened on Nov 1, 2026

• Tests: two cache-prune tests fail under load (a fixed PID 4711 that can be alive; an atime that may move)

  testing
  lanl/bngsim#1440 · Filter by author wshlavacek opened on Nov 1, 2026

• NFsim draws some reaction picks from a process-wide RNG that every prepare reseeds: a seeded session's trajectory moves when another session is prepared between its segments

  lanl/bngsim#1438 · Filter by author elyfmiller opened on Nov 1, 2026

• SSA: recurring events far from t = 0 lose part of a clock spacing each, and a busy run comes back slow with no error

  bugcorrectness concernfamily: SSA exactness
  lanl/bngsim#1436 · Filter by author wshlavacek opened on Oct 31, 2026

• A forward-sensitivity run refused partway arrives as a plain SimulationError, told from a solver failure only by its text

  lanl/bngsim#1435 · Filter by author elyfmiller opened on Oct 31, 2026

• Nightly sanitizers: reports detected

  lanl/bngsim#1429 · Filter by author github-actions opened on Oct 31, 2026
  2

• Leftovers from the #1353 review: a commuting pair on one instant is refused, and a dropped fire is not asked

  lanl/bngsim#1425 · Filter by author wshlavacek opened on Oct 31, 2026

• SUNDIALS: drop the #1301 carry once LLNL/sundials#1073 is merged and the pin moves past it

  gated
  lanl/bngsim#1410 · Filter by author wshlavacek opened on Oct 30, 2026

• amici_sens nightly: BIOMD0000000457 compiles twice on some nights, tripping the wall alert

  testing
  lanl/bngsim#1386 · Filter by author wshlavacek opened on Oct 29, 2026

• Evaluate from_bngl dropping post-generate_network actions: 14 of 103 benchmark BNGL models are benchmarked from seed state, not their authors' protocol

  questiontesting
  lanl/bngsim#1366 · Filter by author wshlavacek opened on Oct 28, 2026

• run_replicates / run_batch: ensembles from consecutive base seeds share almost every trajectory

  correctness concernenhancement
  lanl/bngsim#1365 · Filter by author wshlavacek opened on Oct 28, 2026

• Nightly parity: regressions detected

  lanl/bngsim#1355 · Filter by author github-actions opened on Oct 28, 2026
  6