I probably posted something like this before but it is my understanding rrod (and the similar faults that have since "replaced" it) is mainly a creep problem associated with the ball grid array (what the CPU and graphics use to connect to the board) which is not helped by a badly supported board (the motherboard bends which why a few of the better guides to repairing a 360 will tell you to support it) and weakly held clamps (there is a reason the clamps are replaced) as well as whiskers (
http://www.cinemag.biz/rohs.html ).
Bad joins/bad flow due to lack of lead in the solder (what the heating part of the RROD fixes aim to sort and towel fix is a poor version of) are also present in similar devices.
On top of this you have the usual computer chip issues (electromigration
http://www.dwpg.com/content.php?contid=2&artid=68 and the like) but that really goes without saying.
A returned RROD 360 will often have an extra fan and a couple of replaced capacitors, some read into this that it fixes a fault while others would say litigation dodging (I favour a bit of both and neither).
Of course you also have a near complete lack of a widespread study into 360s (or at least a published one; I would like to think any internal research would be published widely enough to be leaked). A lot of what exists is based on small numbers and inferred from people fixing 360s and similar tech (graphics cards also fail by similar methods).
Heating issues are always present when it comes to high performance electronics but by and large these are minor compared to the ones above and it is usually user error that causes overheating (sticking it in a place without much airflow usually with a justification like "my PS2 worked fine there for years").
Onto pre-emptive fixing; your aim appears to be keep it for as long as possible and resale, if any, will be to someone who appreciates the issues so we are down to purely technical merits.
I would not like to call xclamp replacement; one the one hand it should hold things better, on the other creep rate is proportional to strain (which the new xclamp increases) which may not be a problem worth considering on a good BGA but a bad one (which many are) may cause failure. That is pure hypothesising and I have not done (and to my knowledge nor has anyone else) any tests on 360s or a suitable analogue with a sample size that could be considered good.
Long story short nobody has much of an idea about it all and there is as much pointless discussion and misinformation with good stuff buried in there as there is about bannning. I think I will leave it at keeping electronics cool is almost invariably a good idea and MS decided to either not know or not build electronics using accepted knowledge about how to build electronics, alas games are nice and they have some nice games.