The intricacies of emojis show that they are indeed a whole new language. If you liked the ZWJ mentioned, look into the family or group of people emojis. They are especially complex.
- How long before fonts will start encoding complex parametric curve generators, to generate complex emojis on the fly, instead of whatever they're currently doing (which I imagine is a bunch of precoded vector graphics + code to recolor and glue them)?
- How long before such fonts become some complicated that they become exploitable? I mean, we can already do DOS attacks with Unicode text, but I wonder when it'll cross the threshold to arbitrary code execution?
- How long before encoded emojis themselves become Turing-complete? 10 years from now, someone will want to write: Person Facepalming + Bald + Male Sign + Light Skin Tone + Military Uniform + Red + Pin + Gold, [several similar entries describing the rest of the bridge crew of Enterprise-D], Each, Person, Standing In Circle, Each, Robot, Dancing - and we'll have a loop and a conditional branch, pretty much an embedded Forth.
On the one hand, that last thing would have disastrous implications - on the other hand, we can only cram so much complexity into emojis before someone will start asking for tools to manage it better...
But combining codepoints into graphemes is different than combining graphemes into text. I'm not as clear on the implications of that. Is it really necessary to list out each possible grapheme? Certainly for software to render it, it will need to be able to map them to glyphs in the font, and enumerating them all is one way of doing that, but I can't imagine that being a common approach. I would think that software that cares would add a small special case to handle placement of the individual composed glyphs, and software that doesn't would fallback to showing each glyph adjacent, rather than organized in a block with children in front. There might also be software that assumes an upper limit on how many codepoints can form a grapheme.