Add optimized r_out in stochastic optimization simulation to the output for...

        sample_space[param_name] = np.full(n_samples, float(fixed_value))

    # broadcast locked (non-varying) parameters

    # Use list() to ensure each sample gets an independent copy of list parameters,

    # preventing shared-reference issues when optimization modifies values in-place

    # (e.g. r_out[-1] for COAXIAL pipes)

    for param_name in locked_param_names:

        sample_space[param_name] = [

            getattr(config, param_name) for _ in range(n_samples)

        ]

        val = getattr(config, param_name)

        if isinstance(val, list):

            sample_space[param_name] = [list(val) for _ in range(n_samples)]

        else:

            sample_space[param_name] = [val for _ in range(n_samples)]

    return sample_space, varying_parameters