Hi KarenMarie,
This is a very interesting discussion, so I'll add my 2 cents ...
All my comps [10 PCs running win xp, 1 PC running suse-linux, 2 macs running OS X Tiger (now)] run att 100% CPU load crunching BOINC (10 comps) or seti@home classic (3 comps). The processors range from Pentium III running @ 833 MHz to Athlon 64 3200+.
I have been concerned about temps for a very long time because I have 7 of my comps in one room (which BTW never ever need heating, not even in the harshest of winter days when temp drop to - 35 degrees C). Most of the problems are of the opposite kind.

My oldest comp has been running for 9 years 24/7 before a capacitator on the mobo gave up. I did find a used compatible mobo and rebuilt the system. The processor and most of other components run perfectly ... this is a PIII and the temps are between 62 and 64 degrees C.
My newest comp is the Athlon 64. It has been running @ 100% for more than 1 years, the temps are 60 - 62 degrees C.
I have not seen any problems in any of my comps.
Of course: these comps have also other work to do: one is a file server, an other a multimedia server, one is a communications computer (scans for viruses etc for all the incoming traffic, is the first firewall, switches to modem if the DSL goes down etc), the next is a scanner-server, one has all the electronic books and serves those files, etc). I have tested the rigs extensively, and I have not ever seen more than 1 % degradation in the tasks which are the main function of the comps. BOINC just interrupts its work when the other process use the CPU.
Similarly to you I clean the comps once every 3 - 4 months (I have no cats now - had 2 of these sweet critters before and then I had to clean the comps bi-monthly - cat hair is pesky for comps. The cats just loved to sit on the top of the comps...
I have tried to throttle the cpus using Threadmaster. The temps of the CPU (a PIII) started dropping significantly when I throttled CPU-usage for seti@home (this was before BOINC) below 70%, but the problem was that when I used an other program, which required more than 30% (i.e. the CPU-load became 100%) the program ran more slowly. E.g. photoshop and seti:
Without Threadmaster: seti without photoshop: 100% CPU-usage. seti with photoshop: seti: between 0 and 98% (depending on what photoshop was doing), photoshop 2 - 100%, depending on what photoshop was doing.
With threadmaster: seti without photoshop: 70%, seti with photoshop: seti 70%, photoshop 30% no matter what photoshop was doing!
Thus: threadmaster gives you control of the CPU-usage, but at a cost: you isolate the different processes from each other and when an other program needs more CPU-power it will not get it. I uninstalled Threadmaster ....
Today I am only concerned about CPU-load and temps when I add a new comp (which will not happen soon unless somebody gives me one ....) and the only routines I have is the cleaning out ot the dust (every 3 - 4 months), checking on the comps daily, and rebooting them once a month after cleaning out the temp-directories etc. On the other hand: I had my comps run (just to test them) for 120 days without rebooting and that was OK too.
Now: let me say that this was some 5 years ago, now Threadmaster has evolved (I hope) and may assign CPU-usage more dynamically... but I will not use it because - as I see it - there is no cause ...
A long post but: these are my 2 cents ...
Edit: spelling, addeed last paragraph.