It is not possible to do any kind of test on test net, because the most important thing we want to observe from a change is its economical impact or its value (if bitcoin worth 0, no one cares about how brilliant its technology is, and it will be useless), this can only be done on live traffic

But because bitcoin is the only meaningful cryptocurrency with enough serious traffic, you can only run the test on bitcoin live traffic, which means you have to reach consensus first. A hard fork is inflation and will kill bitcoin's biggest promise of limited supply and will crash its value, thus will not be accepted by any rational participant

So I think the best test is first let 1MB block fill up, and see what happens, how many people complain and how is each one impacted. If there is no real serious impact or people find work around for their problem, then 1MB can stay for a while. If there are some serious issues popping up, then it proves that 1MB is not enough, so you can raise the limit and everyone is convinced

However, if you start to test with a large block size before 1MB limit was reached, then you would never have the chance to test 1MB full block at later time, and if 1MB is really a brilliant and agile solution, you just forever missed the chance to test it