A great deal of functionalities is undocumented despite being around in the same form for a long time.
While many of them are 'documented unofficially' here or on mathematica.stackexchange the fact the they are not officially supported means:
we can't expect WRI Support to answer (though they are helpful more than they need to)
it is very risky to build commercial solutions using them, and very cumbersome without them
I would really prefer that than anything new.
I'm talking about:
Absolute must, how come it is not documented when it is now the standard way of managing/distributing packages?
People use InheritedBlock, WithLocalSettings, HandlerBlock! all the time. They are really helpful
Responsible for dependencies collecting for e.g. CloudDeployed APIFunction/FormFunction. But not only, stuff it contains would be a great help for advanced developers.
Why don't you give us tools as opposed to the solution which does not fit all use cases?
Extremely helpful framework behind Graph's plotting feature (see what we know so far: mma.se.com/q/155237
Stagnation in this area saddens me deeply so I will just make a note that I care :-) And I know that supporting three front ends on different OSs and browsers is a huge task.
There are cases of hard to debug issues that were not fixed since, sometimes, V9. If you are not planning to fix them maybe it is time to talk about them in PossibleIssues section as I understand that they not always are pure bugs but a side effect of subtle optimization patches.
Did I forget about anything? Do you agree?
I thought I'd register my agreement. As someone who enjoys dev work, even super basic documentation without all of the "Details", "Related Guides", and "See Also" sections and crud would be very, very useful. If it's implemented in toplevel code it's one thing and I can spelunk those definitions, but half of the stuff I really want to use is either in the FE or in the kernel so I can't even inspect the definitions, much less read the source.