Inside the function MinimizeRandom, the search goal SolveProblem is limited with the function IloLimitSearch. More precisely, the number of backtracks is limited to the value of numberOfFailsPerIteration, a parameter of the function MinimizeRandom.

ILCGOAL1(SolveProblemIlc, IloRandom, rand) {
  IloSolver s = getSolver();
  IlcScheduler scheduler = IlcScheduler(s);
  IlcResourceConstraint constraint;
  if (SelectResourceConstraint(constraint, scheduler, rand)) {
    return IlcAnd(IlcTryRankFirst(constraint),
                  this);
  }
  return 0;
}
 
ILOCPGOALWRAPPER1(SolveProblem, solver, IloRandom, rand) {
  return SolveProblemIlc(solver, rand);
}
 
void
MinimizeRandom(IloSolver& solver,
               IloNumVar& makespan,
               IloSchedulerSolution& solution,
               IloInt numberOfIterations,
               IloInt numberOfFailsPerIteration)
{
  IloEnv env = solver.getEnv();
  IlcScheduler sched(solver);
  IloRandom randGen(env);
  IloGoal goal;
 
  /* GENERATE AN INITIAL SOLUTION. */
  goal = IloGoalTrue(solver.getEnv());
  solver.solve(goal);
  IloNum min = solver.getMin(makespan);
  makespan.setLB(min);
 
  goal = SolveProblem(env, randGen);
  solver.solve(goal);
  IloNum max = solver.getMin(makespan);
 
  /* OPTIMIZE. */
  IloInt iteration = 0;
 
  goal = IloLimitSearch(env,
                        SolveProblem(env, randGen),
                        IloFailLimit(env, numberOfFailsPerIteration));
  
  while ((min != max) && (iteration < numberOfIterations)) {
    iteration++;
    IloNum value = IloTrunc((min + iteration * max) / (iteration + 1));
    makespan.setUB(value);
 
    solver.out() << "Trying makespan: " << value << endl;
 
    if (solver.solve(goal)) {
      max = solver.getMin(makespan);
      solver.out() << "Solution with makespan " << max << endl;
      solution.store(sched);
    }
 
    else /* ... */ 
  } 
  /* ... */ 
}
 

Each iteration may fail either because there is no solution with a given makespan, or because the algorithm could not find such a solution in the given number of backtracks. By testing the number of failures (the value of solver.getNumberOfFails()) at the end of the iteration, we know whether the numberOfFailsPerIteration backtracks have been consumed or not. If not, the algorithm has proven that no solution exists with a makespan of at most the given value. Hence, it is correct to write the following code.

    if (solver.solve(goal)) {
      max = solver.getMin(makespan);
      solver.out() << "Solution with makespan " << max << endl;
      solution.store(sched);
    }
 
    else if (solver.getNumberOfFails() < numberOfFailsPerIteration) {
      /* ACTUAL FAILURE NOT DUE TO THE LIMITED NUMBER OF FAILS. */
      solver.out() << "Failure with makespan " << value << endl;
      min = value + 1;
    }
    else 
      solver.out() << "Failure from fail limit" << endl;
  }