[racket] A primitive more fundamental than a continuation?

>>Now imagine that we lied. We don't really reify evaluation contexts as
[continuation] procedures but as [continuation] objects with additional
operations for inspecting and possibly mutating them. One could imagine that
continuations are organized as a [abstract] sequence of [abstract] frames:

Now I've become slightly confused by the direction the discussion has taken.

Continuations do appear to be a distinct primitive as opposed to procedures,
and Racket provides the predicate continuation? (see Reference s.9.4)

Racket also provides operations for inspecting continuations, for example
(continuation-marks k) returns the set of marks associated with k. (see Ref.
s9.5).

Is the answer  'Racket can't provide mutable continuations b/c continuations
are implemented like procedures'

or is it

'Racket could implement mutable continuations, but has not'

And, to be clear, I don't have a single reason for why it should provide
mutable continuations.

I'm just attempting to check my understanding of Racket's continuations by
investigating some obvious corrollaries.

Thanks

Zack