Please help me understand the convergence assistant. (Note: I request, at least, NI Multisim support in addressing this issue. Thank you.)

[Euler's_Identity]

The way that simulation works is a set of nodal equations is constructed, and then the variables of those equations are solved, including the variables requested by the user, such as via the scope or meters, or as probes in the various analyses.  That said, if any of the variables of the equations diverges (rather than converges to a solution) there is a convergence error and the simulation fails.  With Multisim v10 (and beyond?,) when a simulation fails, a "Convergence Assistant" (CA) feature becomes active, upon the user's request, to see if any simulation parameters can be adjusted such that convergence can be had.  One of the first things the CA does is it reruns the simulation in order to verify the failure, for the failure should occur again if there is a problem that could be solved, right?

So here's my question.  What does it mean when the CA can't repeat the error so the CA fails to be able to even have a chance to correct things?  Does it mean that there is something unstable about the circuit?  That seems to be a reasonable conclusion to draw from such a failure.  In other words, does it mean that something within the circuit in question has a random (indeterminant) element to it, such that the CA, given the same starting conditions, doesn't arrive at the same spot as the original divergence error?

If this is the case, this random idea, what sorts of things in a circuit could cause this? 

[NavyJay]

I would also like an answer to Euler's question.  Under what circumstances would the Convergence Assintant not be able to reproduce the simulation failure?

[Euler's_Identity]

Thanks.  You are not alone.

But I gotta go

[GarretF]

I'm not 100% sure of my answer here, but I believe Multisim may change the way initial conditions are determined when it tries to repeat the convergence error. I don't think it is something repeated because the behaviour of the circuit has a random element.

[Euler's_Identity]

If it helps, the times I've had the problem I've been using "set to zero" for initial conditions, for I must set up the simulation that way.