Most of the refactoring issues we are seeing are probably due to the fact that we are trying to manipulate a ScenePlugin out of a Scene (the GameManager is not a Scene and holds a reference to a ScenePlugin coming from a Scene that might get invalidated by Phaser 3).
Furthermore, if we want in the future to be able to scale, scenes could be hosted on different servers. Therefore, it makes no sense to have one connexion for the whole application.
Instead, we should have one connexion for each scene.
In the case the reconnect event was triggered on "wake-up", the disconnect event was not sent. But because of a bug in the way we handle the setTimeout, the reconnection to the GameScene never happened after we forced the disconnect scene.
This fixes the issue.
Callbacks for socket.io events were registered each time a disconnect was called, leading to message being dispatched plenty of times if there was several disconnections.
Most of the time, sending a disconnect event to one of the players is enough (the player will close the connection
which will be shut for the other player).
However! In the rare case where the WebRTC connection is not yet established, if we close the connection on one of the player,
the other player will try connecting until a timeout happens (during this time, the circle with the name is displayed for nothing).
So now, we send disconnection event to every body (not only the people in the group, but also to the person leaving the group)
Remvoing messages that were no more used in the new messaging system (the code used to handle the message that sent the position of all users on the front side)
