Thinking about this more, doing a bit more research and seeing that the outputs fail fairly often reinforces the fact that the output devices are undersized. They really should be 15A minimum, 20 better and 200W especially if you plan to use nominally 4 ohm loads. Really, a pair (four total per channel) of 200W devices should be used for good drive with margin into 4 ohm loads. Granted it is rated in the manual for 8 ohm only but there is no excuse for the output protection not working and it failing with shorts or low Z loads.
The outputs could be doubled up, but I'm not sure that it is worth the effort.
Here is a suitable device from OnSemi, 15A 200W but it is not a Darlington so drivers would have to be somehow rigged up:
http://www.onsemi.com/pub_link/Collateral/NJW3281-D.PDF
That the outputs fail further suggests that the output protection doesn't actually work.
Looking at the service manual there are a few things that I don't like. The VAS is undersized, in my opinion, such that I'd expect it to blow if the outputs were shorted and the volume turned up. There is no protection transistor on the VAS which is common when a low power VAS is used. I'd like to see a device with 10 times the rated collector current there running at about 4 times the idle current of the existing design.
Edit: Some protection (current limiting) is provided by R318 in the emitter path of the VAS and it might actually save it when cold, but still it is better to design for worst case conditions.
There is output protection and I believe that it is a latching type that on a single trigger puts the amp into protect mode, this is good if it is set at a high enough current level. I think it is since many are able to run the amp into 4 ohms without it going into protect. The odd thing is that the sense for the trigger is on the output device that is passively turned on and not turned on by the VAS which will provide enough current, far more than the passively driven side, to blow the output and itself from what I see so far. Usually, the protection sense is on both output devices. This might be why outputs fail even with the amp having output protection. It would have been far more logical to put the protection on the VAS driven side. It is quite possible that the VAS fails first and then takes out the output stage. I'd really like to take a look at one of these with a failed output stage.
I agree with William that the air flow and heat sink should be better. If you look at the last Figure on the last page of the transistor specs, it shows that at a case temp of 75 deg C the KEC outputs must be derated to 60W from their 100W nominal rating, so keeping them cool is very important:
http://www2.kec.co.kr/data/databook/pdf/KTD/Eng/KTD1510.pdf