Very Computer

I have the problem that every night at 00:00 my linux kernel produces
the message: "kernel: grow_inodes: inode-max limit reached"

I know how to read and modify the kernel parameters with the help
of the /proc filesystem and I am also aware of the fact that
a cron job could started at 00:00 could be responible (Although
I wasn't able to identify it)

But my question aims at another direction:

On the net I found a text
(http://linux.web.cern.ch/linux/lsr/doc/kernel/proc.txt)
where you can read, that file-handels and inode-handles are dynamically
allocated by
the kernel but not freed again?
Does somebody know whether this is on purpose for efficiency or
whatever,
or if it is a bug?

Secondly - and even more important for me - if the number of used inodes
is bigger then inode-max, does that mean that the system will no longer
work reliable, or is this pre shrink method mentioned in
(http://linux.web.cern.ch/linux/lsr/doc/kernel/proc.txt) something that
will work very well?

By the way, I can increase the values of inode-max and file-max. But are
there any upper limits for these values I cannot go beyond?

Thanks in advance for help or pointers...

Thanks in advance,

Markus Jochim