In a maximization problem which involves many optimization variables, the output of problem/objective value is, say, 1, and one of the optimal variable, say d (the solar panel size cleared), is 3. However, if I add an additional constraint d == 19 to the original problem before, the output of problem/objective value is 19 (which is larger, and should be the “correct” optimal). So in other words, the solver does not find the true optimal? Here I am using SCS solver, with increased allowed number of iterations.

Interestingly, if I try another solver, which is mosek and ECOS, with the default parameters, the output of problem/objective value is no longer feasible when I add an additional constraint d == 19 to the original problem!

all we have to comment is line 181 (constraints += [d== 19.35]). after commenting that out, the problem.value become worse. so meaning not able to find true optimal in the joint clearing?

FYI: To make sure the original problem is indeed feasible, I also tried seperate clearing, indeed the optimal solution in the first stage is 19, and second stage has a solution.

The attached zip is the code to run it.
code.zip