I hear you, but I'd respectfully point out that the "fun" and "cool" in that video isn't about labels, it is about the ability to have multiple geometries in a table, and the ability to have many different ways of generating those geometries, either dynamically or on a one-off basis. The flexibility and dynamism of that allows all sorts of techniques you can't get in Arc, for example
That particular video doesn't really use labels except to create default labels on points, and except to show how labels on areas will automatically move around to show as many as possible area labels for a given view.
Better labeling tools are always desired, but they tend to be very case specific. The best way to get what you want is to say what you want.
For example, take the use case in the thread that inspired that video. How should you be able to create labels that are explicitly positioned at different places for each area, and which are not automatically re-positioned?
There is often tension between automated placement of labels, and the desire to override such automated placement when the machine doesn't read our minds as we would like. How does Arc do that, and how could Manifold do it better?
There are different ways of doing that in Manifold. The video shows just one such way. Another way would be to use style overrides for per-area placement. If that is not easy, at least as easy for someone who has learned Manifold as it would be for someone who has learned Arc, then the key question would be how can that be made easier, and what needs to be done which is required but which cannot be done today.
A related question, good for discussion in this forum, is that if something needs to be done which cannot be done today, how should it be done in a general way, that is not just some one-of-a-kind special case. That's important, because if you succumb to the desire to "just do it" and implement something as a special case, well, before you know it, everything is a special case and with thousands of capabilities you end up with far too many special cases to learn and to remember.
It's better to think in terms of how to create some needed capability using a limited set of broadly-powerful building blocks, each of which can be mastered and remembered and which in combination can be used to do whatever is needed. We need to have that in labels, too, and that takes longer to do than to simply whip up a special case.
It can help to compare how Manifold does such things with how, say, ArcGIS Pro does such things. That can help boil down requests to specifics. How would you handle the use case in the thread in ArcGIS Pro?
I don't think anyone at Manifold has ever said "Perfect. Labels are done. Time for sushi." If anything, it's always been "OK, here are some basic labeling capabilities with which you can do a lot, while labels get further built out." But, honestly, I think that is an understatement. You can do some very intense things today with labels, if you get your head around all the possibilities. That's not saying you can do everything, but what is there today really does cover a lot of what workmanlike GIS output requires.
In fact, some things are easier in 9 than in Illustrator or PhotoShop. An example is this image, which is 100% 9 and fully automated. It's true there is tinker time in such things, and you have to learn to use it, and there is always a fair question how capabilities to do complex / intense presentations could be made easier to learn and to accomplish, but... for all that the capability is there today to do some very useful labeling.
For all that, I absolutely agree there are many ways labels could be improved as part of continuing work on cartography. Not time for sushi, yet. :-) For example, labels that follow curvy lines, better handling of web backgrounds in printed layouts (with better automated handling of scale) and so on. The latter is important for labels because many people do mass production using a base layer that is web served, so the labels that generates have to be scaled correctly.
Also, don't forget that improvements in geometry such as flexibility and greater ease of accessing the power of geometry, also play a key role in labels, since quite often what you want to do is to automate the positioning of labels in a way that classic label positioning engines, even very good ones like ESRI's Maplex label engine, cannot do.
I agree all that has to play together, but it is an ongoing task made up of very many small details from related disciplines, all working together. I guess the bottom line is that if there is anything you specifically want in labels that are not there, say what you want.