I've used L298N for a ton of projects, also used BTS7960 43A and 50A H bridges, and DFRobot bridges. I have never had an H bridge fail.

I fuse the H bridge power input in case I short the output or do something dumb in software, and to provide overload protection for the H bridge/motor for stalled motor current.

The H bridges usually have bridge direction lockouts to prevent users programming a shoot thru situation, tho I'm sure nothing is totally dummy proof.  

Never use a library for an H bridge. Create a direction bit and PWM output to drive the H bridge gate. 

Fuses are a wonderful thing. 

Unfortunately the L298's (and the cheap boards which use them) do not have as much built-in protection.

You can see from an earlier blog that clearly they are failing, can one be sure a fuse would have protected them?

From the photos of them it looks like some have diodes of some sort positioned some distance away. Inductive spikes don't care about a fuse. No decent supply decoupling either in the board in the photo below.

Those boards run warmer too (and might require that heatsink). For those that need it, there are just as cheap and more rugged alternatives.

I too don't use Arduino libraries for H-bridges, but I wrote my custom library to help those who do require it. It's cleaner code in any case, and negligible overhead, so using a good library is often better than directly writing registers. Everyone is not as comfortable writing bits into peripheral registers (it entails knowing the microcontroller architecture, and goes out of the window if you move to (say) a newer Arduino.

Those ancient driver chips were state-of-the-art 40 years ago, they still work in properly-designed applications, but when it comes to cheap hobby boards, I don't think it can be expected of the manufacturer to have made much of an effort.

Regarding the "or do something dumb in software" it would be good if the hardware is protected against that (which it is with almost any decent board).

Failure modes are never absolute unless an autopsy is done and an analysis of the application. :-)   

After having several autopsy's , maybe a common cause can be established. Would doubt normal infant failures. 

I actually have robots and Power Wheels cars (running like an EV) , etc....running for over 10 years on these cheap hobby boards and switching power supplies. 

I'd tend to lean to common destructive failure modes like stalled motors without stalled motor protection, banging a fully loaded motor between full forward and full reverse without ramps, and  intermittent grounds/shorts before focusing on internal suppression circuits, altho cant rule out crappy suppression without ohm checking the boards to the schematic. 

Failure modes are never absolute unless an autopsy is done and an analysis of the application. :-)   

After having several autopsy's , maybe a common cause can be established. Would doubt normal infant failures. 

I actually have robots and Power Wheels cars (running like an EV) , etc....running for over 10 years on these cheap hobby boards and switching power supplies. 

I'd tend to lean to common destructive failure modes like stalled motors without stalled motor protection, banging a fully loaded motor between full forward and full reverse without ramps, and  intermittent grounds/shorts before focusing on internal suppression circuits, altho cant rule out crappy suppression without ohm checking the boards to the schematic. 

Regardless, merely by inspection you can see that the "typical" board in the photo above is not great. For sure people are going to experience failures with it. As well as the problems mentioned earlier, I think more can be heaped on it, just looking at the photo. I can't even see any sense resistors have been installed on it. If a motor stalls, it may well overheat. If you're making a toy robot, you may well be willing to accept the occasional failure if a motor stalls perhaps. But Sean was driving model train apparatus, I presume he wants a bit of reliability.

I view any of these devices as IEC rated components - the lifetime and reliability needs engineered in by the user. In selecting a reliable motor driver, one needs to know ( or measure) stall current, typical motor current, under load voltage, gear ratio. etc... duty cycle, cycle frequency, with little motors those parameters are sometimes tough to get.  They can be checked on a power supply with a load display or battery with a DVM with mA input.  

Apologies if I missed where motor load data was taken. Its luck to size a bridge if the motor load is unknown.