Both versions seen here miss overflow detection. This will make a mistake virtually impossible to catch. A good design always include traps for "things that will never happen". My general approach is to during development to call a routine called "crash" which will turn off the interrupts and loop forever, then with a breakpoint on "crash" you can read the stack to see where the "impossible situation" occurred. The version in the final design depend on the actual system and I usually decide on that after the rest of the program is completed because at such time the ability to do something in a crash situation is clearer.

Have fun,

Erik