Also potentially relevant, here is a look at the server processes for the latest release version (mcserver, minecraft_server.jar) and 18w48b (mcserver2, server.jar) in ps on FreeBSD. Both servers are currently idle, no players present, and no redstone devices or other activity going on in spawn. Compare CPU and memory utilization (3rd and 4th columns) between both servers:
USER PID %CPU %MEM VSZ RSS TT STAT STARTED TIME COMMAND
mcserver 21418 0.0 1.4 11908688 1392200 0 Is+J 22Oct18 2176:37.89 /usr/local/openjdk8/bin/java -Xmx8G -Xms2G -jar /usr/local/minecraft-server/minecraft_server.jar nogui
mcserver2 66316 101.1 4.9 19780004 4924556 5 Is+J 22:21 6:54.67 /usr/local/openjdk8/bin/java -Xmx16G -Xms2G -jar /usr/local/minecraft-server2/server.jar nogui
This machine has 12 cores/24 threads, and the 18w48b server is basically hogging an entire CPU thread and over 4GB of RAM continuously even while idle, while the release version is barely using a gig and no CPU.
I've run 4 minecraft servers on this box for 2 years (3 release, one snapshot). It has 96GB RAM and 24 CPU threads (2x 6 cores x2 HT). I have never seen this issue before loading this snapshot, and none of the other servers are affected still - just the one server running this latest snapshot. Both times it crashed it only had 2 players on it. Do you really think the issue is environmental?
This is happening on my FreeBSD 18w46a server as well, but intermittently. Players will log in (or respawn) and basically be 'in limbo' for 30-60 seconds - can see the sky, and things like rain, but no blocks or creatures at all. After waiting for 30-60 seconds, the game suddenly renders and they can play. Also, when exploring the edges of 'known space', newly discovered chunks sometimes refuse to render in, and it is possible to walk off the edge of the world and sort of 'hang' in limbo. Only way out seems to be teleporting. Did not see either behavior in 18w45a at all despite logging a few hours of playtime on that version. Regenerating the world did not help.
Unable to repro on 1.13-pre1. Looks like this is fixed.
Sorry, been too buried in work to play MC for a while - pretty sure it still happened in 1.12.2. We now have a snapshot server up on 18w22c, will attempt a repro tonight.
Reproed last night, even a short 10-block trip triggers this effect:
Rebooted server, logged in, tamed wolf fought as normal
Loaded wolf into boat, took short trip
Destroyed boat to release wolf, wolf would not fight
Logged out and back in, wolf would not fight
Restarted server, wolf fought as normal again
CPU load is still unacceptably high though - snapshot server still keeps one CPU core permanently pegged at 100% when idle, while the other servers only generate negligible load while idling.