I had been running my server via docker using https://github.com/itzg/docker-minecraft-server, so I figured I should rule out some sort of server configuration that is doing. I downloaded the server jar from https://www.minecraft.net/en-us/download/server/ and ran the exact command from that page. I did nothing else, and let it generate a new world. I was able to go through the portal and back once, but upon trying to reenter the nether I got the massive server lag and client timed out.
I'm running it on Fedora 28, openjdk version 1.8.0_212
For what its worth some more things I've tried, without any luck.
Problem doesn't appear to happen on the same world save (slightly older version, but more or less the same world) in single player.
Problem still occurs when using server with a completely new world save (created new world, used /give to give obsidian and flint/steel). Sometimes it happens when going into the portal, sometimes when coming back, but it always happens. It doesn't always cause a hard disconnect, but it always causes server lag such that I can't break any blocks.
The exit portal disappeared in the new world as well.
I tried decreasing view-distance in server.properties to 6 - no change.
Tried using Java 1.13 instead of the bundled 1.8
Tried leaving java memory settings to default (had previously set it to 4GB)
When I downgrade the server to 1.15.2 and create a new world I'm not able to reproduce, so seems specific to 1.16/1.16.1 where I can reproduce consistently after generating a new world.
I'm experiencing the same problem, so I can provide debug reports for this. Some more details:
Relatively new world, not a hole lot going on yet
Server is running inside of my LAN
Server is completely stock
Client using optifine + shader pack (though I have also tested with completely stock client and same problem occurs)
When I entered the portal the first time, after getting disconnected and logging back in the Nether did load, however there was a lot of server lag and the portal inside of the nether had despawned.
Sorry for the comment spam, been trying to figure out a workaround all day and GOOD NEWS! It looks like the bug is related to the new `sync-chunk-writes` option - setting that option to false fixes the issue in all of my test environments (original world, fresh worlds, fresh worlds sans-docker).
In my original world I had to rebuild the portal in the nether, and my overworld portal teleports me to a non-portal in the nether (where the original exit portal was) but this isn't that much of a problem and I can just rebuild the portal in the right spot.