Arbitrary frames in special relativity Consider this quote by  James J. Callahan in his book The Geometry of Spacetime where he summarizes the conclusions of his chapter on arbitrary frames in SR (page 165):

Once again we find that the radar grid and the rulers-and clocks grid disagree. We have further evidence that in the noninertial frame of an accelerated observer G, no coordinates simultaneously give measurements of a single ruler and clock-as
  they naturally do in an inertial frame. A map of the earth suffers the same defect: Measurements on the map cannot all be
  made proportional to measurements on the surface of the earth.
  No accurate map of (a substantial portion of) the earth can be
  made with just a single scale. On the earth we ascribe this defect to curvature-more precisely, to the fact that the earth is
  curved but the map is not. By analogy, we consider that the same
  may be true for spacetime: Since measurements within the accelerated frames that we have considered are not proportional to
  measurements of the corresponding spacetime intervals, perhaps
  spacetime itself is curved. Our speculations can be summarized
  this way:
  accelerated motions ==> noninertial frames ==> curved spacetime 

The thing that i find very confusing in that quote -- and in that whole chapter actually -- is that Callahan is apparently saying that by virtue of being in a non-inertial frame (while still being in a Minkowski space) spacetime is automatically curved. I've already taken a course in GR and i know that can't be right because Minkowski space has a flat metric. On the other hand Callahan's argument seems reasonable. He's basically saying that, for instance, in a rotating frame you can't measure time and space uniformly like you do in an inertial frame (since v is a function of r and hence time dilation will be a function of r i.e. you can't synchronize the clocks in your frame no matter how hard you tried; a similar effect to how you can't uniformly measure distances on a sphere). A similar question arises in the case of a rotating disc, where the rotating observer apparently experiences non-euclidean geometry. But how can that be? We're still in Minkowski space, the riemann curvature tensor must vanish so why is non-Euclidean geometry present? 
My question can be summarized as follows: Callahan, from what I can tell, is arguing that non-inertial frames imply spacetime curvature, even in Minkowski space which completely contradicts what I've learned before. More specifically I need clarification on the part of Callahan's quote which i highlighted with bold.
 A: 
The thing that i find very confusing in that quote … is that Callahan is apparently saying that by virtue of being in a non-inertial frame (while still being in a Minkowski space) spacetime is automatically curved. 

That is not what he is saying. After considering “disagreement” between “radar grid” and “ruler–clock grid” and noticing that similar disagreement exists between Earth surface  and its maps (and noticing that this disagreement is unavoidable due to Earth's curvature) he makes a hypothesis that such a disagreement would also be inherent for the spacetime (which at this moment is no longer assumed to be a Minkowski spacetime).
Notice his language:

By analogy … may be true …  perhaps … speculations …

It is obvious, that there is nothing automatic about curvature of spacetime.

Callahan, from what I can tell, is arguing that non-inertial frames imply spacetime curvature, even in Minkowski space.

Again, that is wrong summary and Callahan says nothing of the sort. He argues that consideration of non-inertial frames together with examples of curved surfaces suggests (not implies, here we make a hypothesis) that the spacetime may not be Minkowski space but instead a more general curved spacetime (and the name Minkowski is not even present in the quote). 
A: I think that Callahan is probably just wrong. The strongest argument that he isn't seems to be that a monograph on special relativity published by a respected publishing house couldn't possibly be wrong about something so basic. But I have a very low opinion of SR pedagogy in general so I'm not much swayed by that argument.
I haven't read the whole chapter, but I read your excerpt and the unpaywalled first two pages. It's clear that Callahan subscribes to the common view that for each observer, there is a particular reference frame that they must use to describe the world. Those who hold this view also seem to believe that it makes special relativity more subjective that Newtonian physics, even though every argument they make about 3+1 dimensional rectilinear coordinates could be made with exactly as much justification about 3 dimensional rectilinear coordinates. I suppose this is because the human brain has hardwired circuitry for reasoning about 3D but not 3+1D, so that relationships that seem obvious and natural in the former can seem mysterious in the latter.
Adherents of this philosophy also seem to believe that it originated with Einstein, but that's clearly not the case. In his original paper and his early popularization, he always carefully uses phrases like "an observer who takes the moving train as his reference body". The train's motion is specified, but the observer's is not. The observer is simply a scientist who notes coincidences of events – for example, that an object passes a clock affixed to the train at the same moment that both of that clock's hands point to the numeral 12. None of the distortions of vision that depend on the observer's motion (aberration, Doppler shift) affect their in-principle ability to note those coincidences. This seems to have been missed by every single one of Einstein's early interpreters. Today it's the norm to conflate "observer" and "reference frame". If you consistently treat them as synonyms then it's merely unnecessarily confusing jargon, but if you treat an observer as being located at a particular place and also being identical with a universe-spanning coordinate system, then you are going to get into trouble. Callahan does that in section 4.1.
If you think that coordinate systems are as important as that – that every time two people walk by each other in the street, events in the Andromeda galaxy are desynchronized by a week "for them" in some deep physical sense – then it's no surprise that you'd end up thinking that accelerated motion in special relativity has some deep connection to general relativity.

A similar question arises in the case of a rotating disc, where the rotating observer apparently experiences non-euclidean geometry. But how can that be? We're still in Minkowski space, the riemann curvature tensor must vanish so why is non-Euclidean geometry present?

The circumference of the rotating disc is simply not equal to 1 meter times the number of metersticks you put around it. There's no theorem of special relativity that says that you can correctly measure the circumference that way. It's related to the fact that if you put clocks between the metersticks, it's impossible to Einstein synchronize them, and if you replace the metersticks with a waveguide and compare the clockwise and counterclockwise speeds of light, they'll be different. I think that all of these results are interesting, and deserve to be assigned as exercises in every SR textbook. But if they appear to be paradoxical, it just means that your picture of special relativity is incorrect, and probably not internally consistent.
I'm not saying that you shouldn't be surprised by the inequivalence of inertial and accelerated motion. I'm only saying that you should be exactly as surprised by it as you are by the inequivalence of straight lines and curves in Euclidean geometry.
