In certain cases, users may encounter varying time steps and bisections when solving the same subcase of a problem. This variation can lead to doubts about the accuracy of the results. However, there is no cause for suspicion, as the results are indeed accurate.

Some reasons are:

It has been seen with some cases that solver in first inspection do not shows any indication of badly conditioned matrices. But solving a matrix sometimes led to ill-conditioned tangent matrices in resolution phase.
Numerical simulation results turn out to be unstable, since they depend on the machine precision.
This may happen due to machine precision if we are comparing in two different machines.
Presence of friction contact in simulation have also reported such instances, but actually friction presence required a non-symmetric solver