Administrating a mixed network would be hell if I couldn't centralize my logging. The good thing is that there's help out there (like ntsyslog) for creating a common logging environment. You can then filter your logs down using logcheck and being aggressive with syslog_ng. But the best part of having a diverse log feed is the log entries, like:
The data is the error code.
Finding koans like that in your morning mail just makes your day because I like the idea that a Zen master is working at Microsoft.
Seen splunk? My friend at OSU swears by it -- I'm looking into it for our own boxes.
Just checked Splunk out. Seems to be a step in the right direction, re: capturing everything and identifying logical linkings in log output. The problem seems at least two-part: a] derive semantic information from log parsing, and b] create heuristics to detect rising problems.