trying to garner the appropriate specs from diode datasheets does not work so well for me.
The spec that is important here is avalanche energy rating. This spec is not always mentioned or even tested. What this means is how much energy (if any) the diode can eat when breaks down in the reverse direction.
All diodes will conduct if enough reverse voltage is applied. This is called the avalanche effect. If the diode structure is not uniform across the entire chip the avalanche current will flow in a small area causing damage. Zener diodes are designed to operate in reverse conduction but an ordinary diode is not. We try to specify diodes such that they will never see enough voltage to cause reverse breakdown. This is not so easy anymore.
Many of us have learned about stored magnetic energy by being zapped. I learned at a very young age by measuring the field coils of several speakers with a Simpson 260. The current from the ohm meter created a magnetic field in the coil. When the current was disconnected the collapsing magnetic field created a voltage in the coil that found a discharge path through me!
What does this have to do with our tube amps? Well the power transformer is a magnetic coil capable of storing energy in its core, and this energy will be released when the current is abruptly shut off. This energy has to have a path and there is a path through the tube heaters. This has worked well for years. As modern power transformers have been cost reduced the coupling between the windings isn't as good as it used to be and the leakage inductance is higher. This can cause a very short high voltage spike to appear on the high voltage winding (most inductance) as the power switch is turned off. The spike can be less than a microsecond wide and 2500 volts or more! I have measured this on a Hammond 274BX. Vintage transformers are usually much lower.
This effect is magnified when a standby switch interrupts the secondary winding itself. I have seen rectifier tubes arc over when the standby switch is opened. I no longer recomment a standby switch with modern power transformers.
With a big old 1N5408 diode that has several hundred pF of capacitance the spike gets absorbed. With a high speed FRED the spike blows through the silicon zapping the diode. Well it turns out that this is a problem in SMPS's too, so the diode manufacturers have specified the amount of energy that the diode can eat in a short spike. The Fairchild Stealth diodes are avalanche rated and I have never seen one blow.
What happened to my FREDs? At the time the TSE and SSE's were designed the FREDs were the hot ticket. The DSEI 12-12A was a good low cost FRED that worked well in my amps. I bought 100 of them 6 years ago and to this date none of them have failed. Most of the second batch that I bought blew up in short order. Clearly something has changed.
It is common for semiconductor to modify the design of a device during its production life. Higher volume devices are moved to more efficient fabs and transfered to larger wafers. The modified devices are tested to conform to the published specifications. Often large volume customers are given advanced samples of the modified devices to verify performance in their designs. Many FREDs including the DSEI12-12 are not avalanche rated so the modified devices did not violate the non spec.