It was using current time jacobians in a few places and linearizing about the new parameters. This broke Jorge-Luis' lqe problems as the converged state looked to be unstable due the using the wrong Jacobian in the warm start. The linear solver would then essentially have to choice but to fail.
It was using current time jacobians in a few places and linearizing about the new parameters. This broke Jorge-Luis' lqe problems as the converged state looked to be unstable due the using the wrong Jacobian in the warm start. The linear solver would then essentially have to choice but to fail.