After severe errors in the application (out of memory, out of disk space, and so on), the replication can't authenticate (401) against the publish instance. However, you are certain that the password used for the replication wasn't changed.
In the request.log, you see that the replication request results in a 401 status code (unauthorized):
10.02.2011 07:29:34  -> POST /system/replication/content/receive.html HTTP/1.1
10.02.2011 07:29:34  <- 401 text/html 187ms
The search index of the publish instance could be damaged. Rebuild it, then force reindexing of the search indexes.