Apex Network Optimizer 1.7.6

Fixes rejoining a server failing in an already-running client. The first join of a session worked; every later rejoin     was kicked during login with a vanilla decode error such as accept\_teleportation was larger than I expected.     Install on both client and server.

🔁 Rejoins now work

   
• The handshake greeting was being silently dropped on rejoin. The client looked up its connection         through Minecraft.getConnection(), which is derived from the player object and only becomes valid on         the main thread. Since 1.7.5 the handshake is handled on the network thread, where on a warm rejoin it runs before         the player exists — so the lookup returned nothing and the greeting was discarded with no error. The client never         replied, the server waited forever, and the connection died. The connection is now taken from the payload's own         context, which is always valid and always the right connection. Handshake failures are logged instead of being         swallowed.

🤝 The send path now waits for confirmation

   
• Each side used to enable sending and receiving at the same moment, so a brief window existed where one side was         already transmitting in a format the other could not yet read. Receiving is now enabled first and sending only         after the peer confirms it is ready. Until then, traffic stays in plain vanilla form, which is always readable.         This removes the last timing assumption from connection setup.

🧹 Cleanup on disconnect

   
• Pipeline teardown ran from a disconnect event that frequently fired after the connection reference was already         gone, in which case it silently did nothing and left handlers and per-connection state behind. Teardown is now         tied to the network channel closing, which always happens.
   
• A warning is now logged if handler installation finds nothing to attach to, instead of failing silently.