I know, the title of my posting sounds “crazy” but this is what happens.

I solve a MIP Problem (scheduling) using GAMS/CPLEX to optimality. The solver returns the optimal solution (I know that this is the optimal one, I am using a test problem).

Now, when I resolve the problem by fixing a subset of the variables to their optimal values, CPLEX claims that the problem is infeasible.

Has anyone any hints of what could cause such a behaviour of the solver? What should I look for, when rying to fix the problem?

I am sorry I can’t give an example now. The model has become too complex, and I have to admit a real mess.
I will try to sort out what happens. I suppose the problem has to do with the use of small numbers and BigM constraints.