Running a Vanilla 1.8 server and I've tested with hosting it on both Windows 7 and Centos 6.2 to rule out an OS/Java issue.

After a period of time, chunks on the server randomly get corrupted. When a player enters the corrupted chunk the whole server crashes, twice (one from the player entering the chunk and I think the second time from the watchdog timer when it tries to come back up). Visually the chunks look perfectly fine in game and they were chunks that had been generated for a while and used to be fine. Some of them have been built in, some of them have only been mined in, and some haven't been changed, just traveled through.

These could also be chunks that we've ran through dozens / hundreds of times without issue, but once corrupted we can't ever go in again.

Once the player has entered the chunk and crashed the server, it of course crashes every time they join until I can edit their position in their player dat to be outside of the problematic chunk.

I have not been able to discern a pattern to which ones are corrupted, they are randomly throughout the world.

I have seen this issue both on a brand new v1.8 map that is about a week old, as well as an old MC1.2.5 map that we hosted on v1.8.

I did try upgrading the server to 1.8pre4 but without regenerating any of the affected chunks and it continued to crash. I have not tried generating new 1.8pre4 map.

I've attached several of the crash reports. The excerpt included below is the first server crash that occurs when entering the corrupted chunk. It's always a Ticking Entity that is the player with a "java.lang.ArrayIndexOutOfBoundsException" failure. The second crash (Watching Server/watchdog) I consider a symptom of the first one, but there are a couple crash reports from that included since they usually follow the other crashes.

Let me know if there's anything else I can provide.