Flaky behaviour under Twisted 10.1

```
What steps will reproduce the problem?
1. Run the server with twistd following instructions
2. Start 2 test logging processes
3. watch tail -n 30 process.log

What is the expected output? What do you see instead?

The two processes should be interleaving their log messages into the output
process.log. Instead, some random number of messages from one or the other
make it into process.log, then it just stops adding info to process.log. The
number of log entries that make it to process.log varies unpredictably.

What version of the product are you using? On what operating system?

Latest svn as of 11-06-2010, Twisted 10.1, Python 2.7, on OS X 10.6.4.

Please provide any additional information below.

I'm sure this is related to the other open issue regarding logging to the
process.log stopping when the last process generating messages quits. It seems
that any change of state kills the output to process.log including stopping the
last remaining logger or starting a second SocketLogger.

I'm guessing that __init__ being called more than once clobbers the state of
logging by re-config'ing it.

Moving that initialization code out of __init__ solves the problem!

That change has been posted to my github repository.

S

```

Original issue reported on code.google.com by `sstein...@gmail.com` on 7 Nov 2010 at 1:39