Do I see correctly that this is the current version?

def derangements(iterable, r=None):
    """
    ...
    Elements are treated as unique based on their position, not on their value.
    ...
    """
    pool = tuple(iterable)
    xs = tuple(zip(pool))
    for ys in permutations(xs, r=r):
        if any(map(operator.eq, xs, ys)):
            continue
        yield tuple(y[0] for y in ys)

That seems wrong and the combination of zip and eq doesn't make sense to me. Using eq doesn't lead to "Elements are treated as unique based on their position, not on their value." And it makes the zip rather pointless.

The output of

for d in derangements([0, False, 1, True]):
    print(d)

is:

(1, True, 0, False)
(1, True, False, 0)
(True, 1, 0, False)
(True, 1, False, 0)

With operator.is_ (as in my original) it becomes:

(False, 0, True, 1)
(False, 1, True, 0)
(False, True, 0, 1)
(1, 0, True, False)
(1, True, 0, False)
(1, True, False, 0)
(True, 0, False, 1)
(True, 1, 0, False)
(True, 1, False, 0)

Attempt This Online!

Yours is the implementation I was endorsing here; I'll change to use it.

@debruijn, are you OK with this? The other function implementations are good, but I think this is the core.

Yes, go ahead with just keeping the main function. Also go ahead with swapping it to what you intended with pochmann3s implementation (what you already did) - looking back at my comment I don't understand my response of "Sure, I will change that if you prefer that version over pochmann3's faster implementation, no problem." other than me looking at the wrong comment. Sorry for messing that up, also towards pochmann3.

Edit: I had also included in this comment a suggestion/remark that I now see is not consistent with the interpretation of derangements that will be gone for in this package, so I have deleted it. Please ignore it if you have seen it via email/notifications/etc.