LuaTeX comes of age

1) Slowliness. This seems to be (at least partially, but probably in a major way) dependent on the fonts being used. Conversely xetex is incredibly fast with the same fonts. Is there any fundamental reason why LuaTex cannot be speed competitive with xetex? Is the slowliness of LuaTex under consideration now that the goal of reaching function-completeness is reached? I have a feeling that being so computationally intensive, LuaTex won't find much space on cloud based LaTeX implementations such as Overleaf, Sharelatex, etc. (because the same hardware handles way less users with lualatex than with pdflatex). This could severely hinder the LuaTex acceptance.

2) Missing glyphs in fonts. Many widely used fonts miss some glyph sets. Typically, some popular fonts miss Greek letters, which are an important part of scientific typesetting. When using things like Word or Libreoffice, you do not even notice this fact, because the missing glyphs are seamlessly taken from other fonts. Is there a similar mechanism in LuaTex? Can I say "Use this font as the main font, but take the Greek characters from that font"?

3) Conversion of legacy classes to luatex. Is there some "porting" guide? I'd be delighted to see the IEEEtran class ported to lualatex.

4) Full microtype support. Is there some fundamental reason why some microtype features cannot be supported in LuaTex? I know that this is true for xetex, and I wonder if it is only a matter of time for microtype to catch up or if luatex misses some features that pdftex had.

5) "killer" luatex packages. So far, the set of packages using luatex is still very small and, most important, too thin to offer many compelling reasons for using luatex over xetex if you just seek unicode support. I wonder if luatex packages are on development for "fixing" things that so far proved too hard to address seamlessly in tex. I am thinking of things like:

- typesetting on a grid (which improves appearance when printing on thin paper);

- balanced colunms on the last page;

- correct handling and ordering of wide floats in two column mode;

- better support for multipage tables with automatically computed column widths (which incidentally would greatly benefit things like sphynx);

- shaded/boxed text that supports page/column breaks without glitches;

- margin comments that know how to push both the comment and the related text on the next page when the comment would come too low on the page (preferring a shorter page to a page with a broken margin paragraph)

- frame based formatting with frames that flow one into the other;

- figures anchored to the page with text flowing around them (which is in fact a special case of the previous point).

For all these things (pdf)latex has some packages, but all of them have broken corner cases. The ugly of working with current latex is precisely that too often you run in some corner case causing you to spend a huge amount of time to work it around.

I am also curious at how the luatex authors consider projects like "sile" where the premise is that "Most of the development of TeX since Knuth’s era has involved removing his early decisions and replacing them with technologies which have become the industry standard [... so that ...] the parts of TeX that people actually use are 1) the box-and-glue model, 2) the hyphenation algorithm, and 3) the line-breaking algorithm".