A huge function (code size wise) does not have to take a long time to execute.
true, maybe I used the wrong word; however, the meaning should be clear.

How can ensuring proper synchronisation be "dangerous"? If you don't properly plan which sections of code can be interrupted
This comes under the heading of "famous last words"
Running sections of code with interrupts disabled is known to result in latent problems.

Erik Malund said: ....
I say: "if a compiler optimisation
I said NOTHING about "compiler optimisation", I wouldn't, I do not use it. This is the very reason you see my recurring grumble "why do they do this in the optimizer, not the compiler"

Also interestingly I don't believe that KeilI library functions .... I don't mean to "single out" Keil
did you not just do that?

And Keil costs big $$$, so I expect a good result.
When you say "Keil costs big $$$" which, of course is relative, value for money is, in my opinion, a much better yardstick, you need to take into account that what makes a company capable of doing things is not "price per unit" but [price per unit] * [units sold]. The '51 compiler market is not the word processor market and, while I have no idea what the Keil 'production' is, I am sure that we are not talking millions of units.

I totally agree that the Keil tools could be improved in many ways, but being a realist, I think they sell you what you should expect when considering the 'budget' they work with.

Erik