I spent some time debugging this on someone else's server, but unfortunately that only revealed what appeared to be completely random process killing. The system is designed to be resilient, but the debug logs appeared to indicate that sometimes, the script was being killed (or failing to execute in some way) very quickly after it started, but late enough to trigger the initial "I've been run" handler.
There are a couple tweaks for 1.2.4, but I'm not sure how much more can really be done. Randomly not running correctly (with no error logs) is very hard to deal with...