I do agree with you on publishing seeds for Monte Carlo simulations however the argument against it is also very strong. Usually when you run a monte carlo simulation you are quoting the results in terms of statistics. I think it would be sufficient to say that you can 'reproduce' the results as long as your statistics (over many simulations with different seeds) is consistent with the published results. If you run a single simulation with are particular seed you should get the same results however this might be cherry picking a particular simulation result. This is good for code testing but probably not for scientific results. I think by running the code with new seeds is a better way to test the science.