As promised, I have begun to delve into the reporting of coalition troop deaths. I noticed how troop deaths categorized as "non-hostile" by the Pentagon were included without differentiation in most reporting unless one digs into the reporting. Icasualities.org, though it reports the raw numbers without differentiating between one and the other, is thorough enough to provide the information for those willing to look.
Including non-hostile deaths tends to give an inflated perception of the level of violence faced by our troops, though it is also fair to question whether the Pentagon always accurately categorizes each fatality.
Note that even during peacetime the military encounters a substantial number of fatalities. See page 10 ("Table 4") here. The lowest number of casualties among U.S. troops for any given year was 758 during President Clinton's final year in office. Obviously it is not entirely fair to count military deaths that would likely happen regardless of the circumstance of the Iraq War as a result of the war.
The graph, without further ado:
The blue represents total coalition casualties (deaths).
The black represents the number of deaths classified as "non-hostile."
I don't guarantee the accuracy of my count, by the way. My eyes tend to bug out when I count while scrolling. If it's not accurate it should be close, however.
Minus increased operations that result in an increased risk of non-hostile fatalities one would expect a relatively constant rate of those fatalities. The spike in numbers for August and September of 2007 counts as a bit of a surprise. Of course, accidents that result in multiple/many fatalities will result in exactly that type of phenomenon.