as of late the PSU's seem to be the weak point. Also one that should be very servicable.
You are right, someone recently reported a 25% failure rate on s19s PSUs, they seem to have fixed the heatsink/chip solder issues they had with the 17 series and now started to screw with the PSUs.
That is astonishingly bad... Switching power supplies are not rocket science and by nature of their simplicity, should be damn near failure proof. Having a 25% failure rate points to some serious design compromise issues in trying to get the lowest possible build cost. Hell, even a 1% failure rate should be cause for concern.
My 1st guess from a manufacturing cost standpoint would be the output filter caps. Switching supplies operate at a very high frequency - well over 100kHz - and that demands very careful selection of the caps. They MUST be a type with very low ESR (Equivalent Series Resistance) using either a X5R or preferably, a X7R material for the dielectric. Thing is - those cost over 2x what a common general-purpose ceramic cap costs and are produced in much lower volumes than the more common and more easily available ones. With the switching frequencies the PSU's operate at using electrolytic caps is out of the question when designing for any sort of reliability plus would be too bulky compared to ceramic caps.
While you
can use the lower cost common ceramic ones they
WILL fail from internally overheating and current erosion even when using a lot more individual caps to spread out the loads reducing current spikes and resulting heating in the individual caps. At that point it becomes a balance between reliability (use the correct low ESR caps or more of the common caps along with just how many more of them) and cost. Think we know who had the final say: Accountants, not the Engineers...
The line side of the supply is harder to screw up by using under-spec'd (cheaper) components and same for the switching FET's, we're talking only a penny or two difference in cost between parts rated for x voltage/current and ones rated a bit higher. Again, it would be great for someone to do forensics on a few dead PSU's to find out what their failure mechanisms are.