On moving forward and breaking compatibility
is a gentle rant about precisely this. We had about 2 years of "crap, my recovery script I run once every forever is dying and $APPLICATION is dead in the water! HALP!" emails at $PREVIOUS_WORKPLACE
because of this change. The old behavior would allow these scripts to work because autocheck was off, and the scripts were checking the errors; once it was on automatically, the scripts would crap out despite the "proper" error handling being in place.
The middle of an emergency is not when you want to try to explain how this change that has broken things and is costing money was, really, a good idea, and how, yes, they will have to patch the script if they want to move forward, and yes, they will have to do this on the production machine, and yes, they will have to get all the relevant people out of bed, and no, I don't think that's where the author's head was.
As I mention in the above node, this could have been done better, with examples of how so.