It all depends on what you deem to be reliable. Making devices more 'reliable' generally involves extra cost. When designing a product, one tries to cope with any forseen problems but even then you sometimes have to debate the use of protecting against a probable but unlikely circumstance.

For an example: If I design an alarm system to be installed in commercial premises - I design in protection for the inputs and outputs so that they will withstand short circuits, short to 12 volts, ESD and they don't radiate excessive EMI nor are affected by a moderate amount of RF pickup. This should give me a reliable product in it's intended application. In the real world some monkey installer connects 240vac into the 12vdc for the remote panel. Smoke ensues and the panel fails. Should I have designed in protection for this eventuality? Commercial sense says no as it would cost too much.

Refering to the power supply lasting 300 years based in the MTBF, I doubt if that would match reality. MTBF is a notional value based on the statistics of the individual components. I dare say the power supply would last about 5 years max at 60% load at moderate temperature due to the electrolytics drying out. Even less at higher load and/or temperature.

I'm sure they could design a car that doesn't need a whole heap of things changed at 100000kms - but would we want to pay for one?