Put that in cron for future reports avalibility
Syslog should be your first look (e.g kernel will kill processes which hit the kernel limits, and write that in syslog )
Other then that, if you are using Nagios,HP OpenView/ SIM or alike, some data should be there.
If you want to catch 'what is wrong' you need to constantly monitor your HW and use monitoring tools like Nagios/OpenView/HPSIM, create rules and such.
Example, you use sar in cron to catch in what part of day/week/etc. the problems are occurring.
Presuming the time, write script / configure monitoring tool etc. to run some command when machine is constantly 100% or alike..