This topic may be elsewhere on the board and I'm not finding it, so please excuse me if that's the case. My question is about the new material in Article 130.5 (A), specifically the term "likelihood of occurrence". What precisely is intended to be used to determine this? That term feels like it's intentionally ambiguous.

It's not really. Certain tasks are clearly going to have a very high likelihood. For example human error rates obtained through a lot of different studies reflect an error rate of between about 1% and 40%, which is vastly higher than the typical 1 in 100,000 to 1 in 1 million or lower rates required. The actual minimum requirement is jurisdiction and industry dependent, and a little hard to find.

When the task is primarily determined by the reliability of the equipment itself, IEEE 493 is the reliability standard and shows that equipment reliability meets those standards. There are some others such as OREDA as well that give various data points. That being said, that's an AVERAGE under industry specific conditions and doesn't reflect local conditions. With mechanical equipment, reliability rates are roughly 60%+ determined by the environment and nothing to do with the equipment reliability. So any data "book" is kind of useless and actual field experience is necessary with mechanical equipment. Since we have so many specifications on the electrical equipment environment, that avoids the problem.

Finally there are some field determined conditions where obviously we can simply ignore what the book says and take appropriate action. For example if a breaker or fuse recently tripped, we know for sure that there is an electrical fault which clearly raises the possibility of a serious issue. Another example would be soot marks, smoke, or liquids all over the surface of the cabinet indicating that something is either ongoing or has recently happened which significantly increases the likelihood that something is wrong...no telling without opening the cabinet how bad it is but no doubt that this changes the scenario significantly. Otherwise as stated in NFPA 70E itself, equipment which is properly maintained and properly installed is unlikely to pose a significant hazard under normal operations which would obviously include routine maintenance tasks.

There are standards for determining this under both IEC and some ANSI standards for determining risk assessments which includes equipment reliability data.

The easiest source for "likelihood of occurrence" of an arc flash incident is Table 130.5(C). Then couple that with the incident energy from an arc flash study or from the 130.7(C)(15) tables at that particular piece of equipment to get severity of incident relative to injury. For example if the equipment has an IE less than 1.2 cal/cm2, then possible injury is obviously not as severe as equipment that has an IE of 4 cal/cm2 or higher. From 130.5(B) also take into account operating condition and condition of maintenance. For example, if there's no obvious signs of problems versus something with burn marks, smell of ozone or burning, hot to touch, smoke, etc. Take those three together to get the "likelihood of occurrence of injury or damage to health and potential severity of injury o damage to health".

This contains a chart that probably shows what you want:

http://ieeexplore.ieee.org/document/6164469/

The one exception in this table is that Roberts suggests that medium voltage equipment is inherently risky (or at least more risky than low voltage equipment) which is ridiculous and baseless. It's something I see all the time with people that are basically totally unfamiliar with medium voltage equipment.

Also you may want to look at the annexes in OSHA 30 CFR 1910.269 which has a much better table than the poor one in NFPA 70E. It's utility-specific but does a great job in a very small amount of space, particularly describing what "abnormal" or suspect equipment looks like by giving a specific list instead of some vague terms about the equipment condition.