I just installed SA on a stock CentOS 5 machine. I forwarded a spam to it
and it worked. Then I sent a very short "hi" message that I would expect
should make it through but it did not.

From looking at the X-Spam headers:

X-Spam-Status: Yes, score=5.3 required=5.0 tests=AWL autolearn=no version=3.1.9
* 5.3 AWL AWL: From: address is in the auto white-list

I don't see an explaination. Where is the explaination?

Why was this flagged?

Is SA taking into consideration that I previously sent a spam from the
same address?

Was it flagged because the message was so simple?

Why did AWL *increase* the score rather than reduce?

How does SA know that the address is in the auto white-list?