Evil!

Since it's a computer markup language, and defined by your standard-issue Context-Free Grammar with recursive productions, you can wind up with arbitrarily complex blazons and make some really neat-looking stuff. See also PyBlazonHacks.

• Or three groups of three groups of three groups of three groups of three shakeforks sable. This one has been mentioned elsewhere, but it bears repeating. Two hundred and forty-three shakeforks (why shakeforks? Because they look good with all the other threefold configurations happening here).

• Argent semy of roundels sable semy of lozenges or semy of billets gules. You probably want to stop this after the lozenges or (as I did), but yes, you can color the charges of your semy with tinctures which have semy charges themselves, and so on. And here the power of SVG shines forth, since you can actually zoom in close enough to see even the innermost charges. But be warned; doing that can seriously munch your processing power for a while. Depends on the library you use, I guess.

• Azure semy of roundels argent fretty sable, compare with Azure semy of groups of 2 roundels argent fretty sable, and, say, Azure semy of groups of 8 roundels argent fretty sable 3 2 and 3. Note how the scale of the fretty is fixed with respect to the semy arrangement, and doesn't scale down to the individual constituents. So each of the roundels in the groups of eight barely holds one diamond, while the ones in pairs hold more, and the singletons more still.

• Argent semy of groups of five increscents azure in annulo. Just looks really cool, with those little rings of crescents.

• Azure five groups in annulo of as many crescents argent, compare with Azure five groups of as many crescents argent in annulo. In one case, the groups are in annulo and the crescents in default configuration (quincunx); in the other it's the crescents in annulo and the groups in a quincunx. Similarly, you can specify both arrangements, e.g. "5 groups in annulo of 5 crescents in pale" vs. "5 groups in pale of 5 crescents in annulo," or even "five groups in annulo of as many mullets of as many points in annulo."

• Gyronny of eight sable and argent on a roundel gyronny of 16 argent and sable a roundel gyronny of 32 sable and argent. Just looks neat.

• Sable six escutcheons per pale gules and azure in annulo each charged with a roundel counterchanged. As of the current writing, the "per pale" division of the escutcheons is per pale with respect to the shield, not with respect to each individual escutcheon (since they are rotated around in their annulo arrangement). Is that correct? Probably not, but it isn't obvious. Something to think about. Same problem if you charge the escutcheons with charges arranged in fess or pale or bend or whatever.

• Sable two flaunches or each charged with three roundels per pale vert and gules. Because flaunches (as of this writing) are done as reflections of each other (since they always come in pairs anyway), you will find that the roundels on one of the flaunches are actually per pale gules and vert (i.e. mirror-image of what it should be). Is this correct? Again, probably not, but an interesting question.