Today i have found the Scripts part of the Wiki.
There is an interesting entry about STRING declaration.
The author is EvilSOB and he stated this (for local STRING variables):
Is it true for the latest A7?
What about with other local variables, var, int etc.?

This is not true for global strings that are defines outside a function like this:
STRING* my_string = "Hello!";

However, if you create your string with str_create within a function like this:
STRING* my_string = str_create("Hello!");
then, you'll have to remove it with ptr_remove at the end of the function if you don't need it anymore. Otherwise it indeed stays in the memory and the memory consumption of your game increases each time you call that function.

Its still true with the latest A7, and will probably always be the case.

After all, this IS NOT a flaw, it is most often a good thing...
Think how difficult it would be if you could create entities in
a function, like bullets, but they cease to exist as soon as
that function finishes...

It is only applicable for objects that you manually create using an
engine 'create' function. EG ent_create, bmap_create, str_create....
(OR sys_malloc if you are ready for that yet....)

Thanks, as i wrote i ment local string.
So, thank for the info.
What is the situation when the acknex is crashed?

Yeah, I figured as much.

In the case of a application crash, sometimes all the memory gets
cleared, sometimes it doesnt.

But thats more or less up to how well windows handled the crash.

It depends a lot on what type of illegal operation it was that caused the crash...
If windows got upset about what acknex was doing, it can be bad.
If acknex saw a fault and crashes out, its usually pretty clean.

Ive had as much as 750M of ram locked out after a bad 'malloc' command.
But EXACTLY the same fault when using 'sys_malloc' and I lost no memory.
Thats because (as far as I can tell) the malloc got caught by windows,
whereas the sys_malloc got caught by acknex.