Not sure how I feel about it. I can see the former being useful on smartphones, but I really only have the latter available on the smartphone.
I think the current behavior of keyboard/menu Zoom is meant just as an aid to people with poor eyesight who can't read small text. The change from Text Zoom to Zoom is just because increasing the text without increasing anything else can cause layout issues, so it increases everything. But it keeps the document width the same because scrolling horizontally to read text sucks.
Meanwhile pinch-to-zoom scales everything because it acts just like iOS, and because pinch-to-zoom isn't supposed to ever reflow anything (as that would be inconsistent with pinch-to-zoom everywhere else). So pinch-to-zoom is useful when you want to quickly zoom in on an element, as opposed to increasing the text size of the entire page.
All that said, it's still interesting that keyboard/menu Zoom doesn't handle layer scaling correctly. I'm guessing this is because the Zoom is not, in fact, scaling the whole page. It's reflowing. I could imagine that if it simply scaled all transforms you might get some odd behavior in various cases where it escapes the bounds of the document. That said, the current observed behavior is even stranger. It's hard to imagine what exactly is going on with that transform as I adjust the keyboard/mouse Zoom.
I've used the matrix function a lot, and am perfectly happy with how it renders across browsers. I would be happy if it was bug-free and easier to use, but it still gives me a lot more power than if it didn't exist.
Since about 2011, the pace of change has been enormous. Concepts which we only dreamed about in 2005 - transitions, animation, flexbox - now have a spec and are even supported. OK, some of the implementations are buggy, but the alternative would be for browser vendors not to implement them in production until all the bugs were ironed out.
I'm happy that the browser vendors release code that is largely working, despite the bugs. I have confidence that they will fix those bugs. If I'm happy to use the buggy implementation, I can. If I prefer to wait for a bug-free stable version, I can. The choice is mine.
The sense of entitlement amongst many front end developers is immense. They contribute nothing to the development of either the standards or the browsers, and yet they moan when it takes more than a few months to introduce a completely bug-free, consistent browser implementation.
One would expect browsers would at least get the _MATH_ right.
Yes, in an ideal world, all of your transforming _MATH_ would be in the same place and interoperate. But the reality is that browsers are large, complex systems who's requirements are constantly changing.
Do you think these edge cases deserve a massive rewrite? (which in the short term would probably introduce more bugs than it fixes)
How should browsers handle transitions of transformations?
AFAIK in Firefox if your transformation is a composition of multiple transformations then it does a transition of each part in a way where the transformation parameter linearly changes. Like between scale(2) rotate(0) and scale(4) rotate(90) in the middle of the transformation there is scale(3) rotate(45). It's different if you define the same transformations at the two ends as matrices.
It's kind if cool tho, since you can create a transition between rotate(0) and rotate(360), however it's not entirely clear what happens if the two ends have different transformation elements.
I take it that question is roughly: given two matrices A and B, guess which continuous path M(t) of matrices with M(0)=A and M(1)=A the web developer intended or will find most visually pleasing. I don't think linear algebra has any mind reading built in.
By the way, why don't browsers just let you specify the function M(t) to be used for the transition?
The shortest path depends on your metric. There are straightforward choices for translation (uniform L2 norm) and even scaling (log) and rotation (the amount you rotate by), but when all three are combined in one matrix, units come into play, and things like skew and perspective make them more complicated.
This path is not guaranteed to be unique. Uniqueness fails exactly where two or more geodesics cross. The set of places where this happens is called the "cut locus". The example given elsewhere of a 180 degree rotation is a good illustration of a point in a cut locus.
The set of invertible linear transformations is also not connected. It is partitioned into two components: those with positive determinant and those with negative determinant, and there is no continuously differentiable path from one to the other that does not cross through a matrix with det=0 (i.e., that is not invertible). http://dbaron.org/css/test/2010/transition-negative-determin... has some examples of animations between 2D transformations where the sign of the determinant differs.
As for intent... this is the reason for the rule pornel gives in another comment: if all of your transforms use the same functions in the same order, then interpolation is done on the arguments of these functions (this also eliminates the problems with having to choose a metric described above). That gives the designer a reasonably simple way to express intent.
I think that a language for specifying M(t) that is expressive enough to cover all the bases would be very complicated to define, and probably not of much more practical utility than just defining a large number of matrices and doing short transitions between them, which you can do today.
P.S., I work for Mozilla and gave advice to people in the CSS WG when 3D animations were being discussed.
 The definition of canonical being, if you locked 10 mathematicians in a room and asked them a question, it's the answer 9 of them would give you.
You can't connect a matrix of positive determinant with one of negative determinant without going through non-invertible matrices. (EDIT: I somehow missed that you made the correction about disconnectedness yourself in a later paragraph.)
Even restricting to matrices of positive determinant (the space of which is connected), I don't think there are always geodesics between any pair of matrices. For example in the 2x2 case, I bet there is no geodesic from I to -I.
No no, the stuff from the mathematicians probably wouldn't work and we'd be left with a dent in our egos.
If all your transforms use the same functions in the same order, then transition is done on arguments of these functions.
If functions or their order differs then everything is flattened to a matrix and then the matrix is transitioned.
Edit: Sorry, linear relation is not what I meant - one should be able to construct a parametric equation in terms of t where t = 0 is the initial transform and t = 1 is the final transform.