Many theories have their validity, but the important thing is that the final result must be usable.
Due to problems already at startup of MP sessions, and due to the hard limitation that backward paths are not yet managed in the new MP OR versions, our group is still using Rel.1888 (the last one before Rob's reworking) for MP sessions, and I must say that such release together with the last memory management patches is quite reliable and usable.
In parallel I (and maybe another person) will have some personal debug with a server and a client OR instance, to provide to Rob inputs to get as fast as possible a playable and enjoyable MP version. Up to that moment (that is up to getting hands on experience in MP sessions with the new releases) I won't take positions on theories about what should be allowed and what should be forbidden.
As Rob said that my previous test report based on a Bernina test case was too complicated as a first case, I tried a simpler case, see picture below:
I first successfully prepared the station exit path for my client train, and then tried to clear the first signal in the path. The signal remained at stop both in the track monitor as in the dispatch viewer, as can be seen. It was instead possible to clear the second signal, but of course without the first signal cleared the train can't exit the station.
I tested same situation with release 1888, and here is the outcome:
The first signal on the path is in APPROACH_1 state already at game start. The second signal can be cleared, so the train can exit the station.
If Rob needs further info about this case I can provide it.
I foresee to provide further test cases.
I understand that Rob has a big work to do, but nevertheless I hope that this new MP management will become usable in not too long time, else we will have to stick to release 1888 for MP sessions, even if we can't enjoy improvements in other areas of OR.