Hi,
I have a server (gentoo) which has been running for 2310 days without restarting. About a few months ago a weird segmentation fault error started to happen when we ran perl scripts and ssh client. I reinstalled new version of perl and new ssh client not linked against openssl and we stopped to use old binaries. This is how we "solved" it. Month after that I ran ssh client and it worked, so it appears that the problem was fixed "itself". However a few days ago an oracle sqlplus binary started to throw the same segmentation fault error again.
I am suspecting that the error is somewhere in network libraries but I don't know what causes it. I have tryed 'strace'ing it, but can't find any clue. The server is our production environment and I suspect restarting it might fix, but I don't want to restart it since I don't know if it will start up ok, it is a very old machine.

my question is, is there anything I can do to somehow clean the disk cache of the server without rebooting it? Maybe a library file was corrupted in memory and by reloading it I could fix it.
What other things I could do to discover what causing this segfault?

Any software update over the last 6 years may have changed a file (such as a library) on disk but not the old copy in memory. No easy way to fix that other than reboot.

I have not made any updates to the distribution since the installation. All the new software goes into /usr/local and oracle in /home/oracle so what you suggest doesn't apply in this case.