Fair enough. As you say, copper is reasonably dense, and certainly toxic in a sufficiently large dose. Part of the reason why lead has a worse reputation is that it isn't as easily eliminated from the body, so repeated small doses have a cumulative effect. My point was just that that, as far as crackle is concerned, I wouldn't interpret the term 'heavy-metal-free' to mean that there is no copper oxide present.
The idea that metals with more than one stable oxidation state are involved certainly has merit. It was part of Shimizu's original suggested mechanism and, while I don't buy the idea that preferential consumption of Mg causes the smoulder phase, partial reduction of the metal oxide can't be eliminated so easily. Unfortunately, as with many of the proposed mechanisms I've come across, it can't be the whole answer. About 4 years ago, when I first started getting interested in how crackle worked, I made a series of tests of approximately stoichiometric mixes with varying ratios of copper and bismuth oxides. They all worked to some degree or other, including the one that contained 88% Bi2O3 and 12% magnalium (plus 10% NC). A 2mm grain, when ignited with a blowtorch, produces a series of 8 to 10 sharp reports, with virtually no delay. It isn't the best of crackle compositions, but it does work.