$\def\ns#1#2{#1_{\rm#2}} \def\xA{\ns xA} \def\tA{\ns tA}
\def\xB{\ns xB} \def\tB{\ns tB}$
I'll try to answer your new questions.
Here you have Alice, the state of which is known and the space time is
kind of fixed to her frame...?
Not at all. it's better to think of spacetime as independent of particular observers and their frames. Maybe an analogy with Earth's surface helps.
Since more than a century international agreements exist as to which
geographical coordinates to use: latitude and Greenwich longitude. It was not so before. French used Paris meridian as origin of longitudes, Italy (I suppose) Rome's meridian, and so on. (Latitudes were uniformly reckoned, since poles and equator are privileged references.)
The same, and more varied, happens in spacetime. Every reference frame is as good as any other, thus spacetime coordinates used in different frames are equally valid. Alice will use $(\xA,\tA)$ coordinates, Bob $(\xB,\tB)$ (but with an important proviso, see later) etc.
The proviso is that only inertial frames are allowed and this causes
a difficulty with Bob and Carol, who change their speeds halfway. Two different frames must be introduced, both for Bob and for Carol, and a junction must be defined within each couple.
Those spacetime diagrams would be very awkward, as a consequence of what I said before. So much so, that it would be better to invent a "virtual" Alice with its frame. We are always allowed to do such things, since physically relevant results, relating to measurable quantities, are to be computable in whichever frame we like. In physicists jargon, are invariant.
2/2 Case 2: Suppose we also have Alice but this time Bob is home clock (the home clock in twin experiment).
You already know the answer: this is not allowed. Or better $\dots$ contrary to a diffused opinion, it is possible to use accelerated frames in SR. But at the price of considerable complication in reasoning and calculations. You ought to model Bob's motion wrt to an inertial frame, then to introduce special coordinates for Bob's accelerated frame, and so on. Out of reach for beginners, as well as useless.
In my opinion, for this and other reasons, it is better not to think in terms of time dilation (and to leave apart Lorentz transformations, seldom needed in elementary problems). This is why I suggested to work with spacetime paths and their lengths. An approach leading you to think of the physical situation and the quantities requested, instead of relying on rote formulas, prone to frequent misuse.