It is funny that you should say that Bridge rectifiers are robust and don't fail very often. The chargers on my wife's wheel chairs (she has two identical units) were failing at a high rate on average lasting less than a year, I ended up buying them in tens, went through 15 in seven years with the two chairs. The chargers are rated at 24 volt 10 amps the rectifier is a GBPC3504 rated at 35 amps 400 volts. I was replacing with the original parts. I then ordered a batch from CPC Farnell and since then there has been no failure. But now the chars have been replaced and the new chargers don't seem to have the same problem.
That is a massive failure rate indeed!
This does sound like poor cooling design to me because they fail so consistently. The specs of the part itself certainly appear to be sufficient. It will dissipate roughly 20 Watts in a configuration like this.* The design might have been dependent on heat-tolerant parts. Some products are more tolerant of over-temperature than others.
*I usually estimate the power by multiplying the average current by 2 assuming a 1 Volt per diode voltage drop, the exact calculation depends heavily on the voltage source impedance and the capacitor ESR. It certainly is a quite rough estimate and is likely to be on the high side.
I have seen passive PFC where it is a massive LC resonant circuit in series with the one mains lead, where the LC is tunes to 3 times the mains, to trap the third harmonic, and as a bonus it increases the rectifier on time while reducing the distortion in the mains.
I have seen bridges fail, 4 diodes are generally more reliable. I have a ferrroresonant ups that draws a near constant current, but which has a near unity input power factor, though the output is slightly distorted, though a sine wave, as it has a 100uF output capacitor.
Love these "spare-no-expense" designs! I usually only see the cheap-ass consumer goods.
Separate diodes can be more reliable, but usually only due to the lower volumetric power density and greater thermal inertia, but there are exceptions:
Electric designSome designs are just poor, but others can evolve very unpredictable problems that can only be discovered when tested thoroughly or when they fail "in the field". Problems with inductive kick-back from the transformer does exist and can be found when load currents rise and drop fast and only very limited storage capacity is available. Electro-mechanical equipment like DC motors and switches (these need not be mechanical) can cause significant voltage peaks too.
Over-current is usually much less of a problem because a bridge can handle huge surge currents and a fuse will usually blow before the bridge gets destroyed. Over dimensioning bridge rectifiers is also commonly seen as they are quite cheap.
Thermal designBridges, especially in high current environments can dissipate substantial amounts of power. Bridges are usually built in housings that can dissipate a lot of heat on their own. When convection is obstructed this dissipation mechanism stops and can kill the bridge. Sometimes the bridge is screwed onto a metal part of the housing it is in, but these metals often have very poor heat conducting capability so the cooling of the bridge is substantially less than immediately obvious. Absence of heat conducting paste can also cause a bridge to overheat and is surprisingly common.
Cheap suppliersSome no-name or fake semiconductors get built to very poor construction standards and are sometimes incapable meeting their specs. When buying from respected sources like Digi-Key or Farnell it is hardly likely you will ever encounter one. Most are perfectly reliable, but have alternative issues like a rough finish decreasing the contact area and subsequently the heat transfer.