Inconsistent timing in runs and between runs using different dt

[Porthmeus]

First of all - thanks for the work!

I have played around with the dynamic parallel FBA function and get overall reasonable results with a 3 member community. However, the time steps are inconsistent. First of all if I run a single simulation time steps change during the simulation:

ds.simulate(0.1)
ds.get_time_points()
# [0,
# 0.1,
# 0.2,
# 0.23333333333333334,
# 0.33333333333333337,
# 0.3666666666666667,
# 0.40951541623750076]

I assume, that you use some dynamic adjustment of dt depending on the FBA results? Can you explain what is happening there?

At the same time, if I simulate the same community with different dt values, I get different values for the last time point. While simulation results are similar enough that I think everything goes right there, I am confused about that:

total = []
for y in [0.005, 0.01, 0.05, 0.1]:
    ds = DynamicParallelFBA(models, [initial_biomass]*len(models),metConc)
    ds.simulate(y)
    x = len(ds.get_time_points())
    total.append(ds.get_time_points()[x-1])
total
# [0.1730408228043126,
# 0.18789289829725145,
# 0.2800212029191405,
# 0.40951541623750076]

I would have expected that the simulations would stop all at roughly the same time (even assuming some dynamic adjustment of dt). Maybe the stopping criterion is different with different time points? (BTW what is the stopping criterion - unsolvable FBA for one of the models?)

I can provide a script and models to reproduce the behavior, however, I doubt it specific to my case. Could you explain what is going on there and whether all of this is expected behavior?

Thanks in advance!