Asterisk needs machine reboot after a while

We have the following installation:

Intel Server with 2 x Quad Xeon, 16 Gig Ram
1 x Wildcard TE122 Card

CentOS 5.2
Asterisk 1.4.3 (same thing with 1.4.22 or older)
Libpri 1.4.9 (same thing with 1.4.7)

The configuration files are created with FreePBX 2.5

There are about 100 Users on this system with 700 calls a day.

Everything is running fine for about 2 weeks.
Then the systeme starts to behave strange: the busy signaling don´t function, agents in queues are getting calls (when they are talking)

First I tried a dialplan reload in the asterisk console.
-> The asterisk still behaved strange.

Then I stopped the asterisk server, I stopped the zaptel. I started zaptel and asterisk.
-> The asterisk still behaved strange.

Then I did a reboot of the complete system.
-> the system comes up and runs fine (for a few weeks).

tia florian

In much more weaker system I had similar problems.
I found some of scripts were memory killers, and some scripts died and became zombie.
Some others detach from Asterisk and consumed huge amount of processor.
All scripts were cleared from issues, and now Asterisk works much longer. I suppose there are few issues not fixed yet and this could be reason.
If you use much of AGI scripts, with lot of while and for - then put everywhere where is possible checking of channel and exiting.
If you use much of sox or some other tools to mix waves - they can do lot of harm also if used not correctly.
If you use text to speech - everywhere where is possible - generate waves and use them.