I also think this method is sometimes useful as I have used it myself. The main advantage I can see is that it doesn't require a complete system reset, but allows "graceful recovery". I mean, if the watchdog kicks in, the uC is reset and has to begin from scratch, losing all information about the state it was in.

If the watchdog is interrupt-based, only the crashed part of the code has to be handled, and it can start functioning smoothly without having to go through a total reset.

Another advantage is that you don't need an external watchdog IC. Also I don't think maintainability will be much of an issue if you use a well-defined way to "register" the watchdog trip handler routine.

But then the issues about interrupts being NMI and all have to be considered to determine its suitability to the application.

kundi