With the awareness of arc flash, many giant manufacturers do not manufacture the Service Disconnect and the Automatic Transfer Switch located in the same section or enclosure. However, this practice can be seen in the field for switchboards rated as high as 600 Amps.
The dangerous part is the upstream of service disconnect is like a blind spot as the only protective device is the utility's fuse on the primary side of the transformer and often result in high incident energy (greater than 40 Cal/cm2 in most of the cases) at the service disconnect. But because of service disconnect as protective device, in the downstream the incident energy on the ATS(normal-utility side) gets reduced to for instance less than 4 cal/cm2. The problem is although ATS has lower incident energy, it is located right below the Service Disconnect in the same section (enclosure). This is a arc flash hazard and I affix the conservative label (service disconnect) on the section that has service disconnect on the top and ATS at the bottom. So please share your thoughts on how you affix labels:
1. When the Service Disconnect & ATS is located in the same section (enclosure)
2. When there is a barrier between Service Disconnect and ATS located in the same section (enclosure).

1. Depends on how substantial your barrier is. If we accept that no testing so far has shown arcs propagating for instance from one switchgear section to another then if the construction of the equipment is the same...that is there is a substantial thickness of sheet metal between the two cabinets, it seems like propagation from upstream to downstream and vice versa is not likely. In that case, you need two labels. Otherwise you just label it for the biggest danger in the whole panel and be done with it.
2. I've run into this all the time. The label does NOT necessarily mean the conditions at the time of performing maintenance and should not automatically be treated as anything other than worst case. For insance 70E-2015 already contains provisions for not wearing arc flash PPE for several typical use cases.
3. I've run into the flip side of this several times where due to the fact that the generator fault current is so much lower, it increases opening times and makes the ATS and downstream equipment much more of a problem.

I agree with Paul with one exception. While propagation has not been seen in electrical equipment between sections it has happened. Where I disagree is in stating 'if the structure is similar...' we can assume no propagation. I am an electrical engineer, not a structural engineer and I am personally not willing to make that assessment.

I would label any single compartment with the worst case. I would most likely label adjacent compartments the same way - worst case. Either would be based on the high side fuse.

Then you have to label it all that way unless it's physically in different gear (ie, separated by cables). What has happened so far in "barier" type testing is that it usually stops at the barrier and the plasma pools there, but sometimes the arc reignites again on the other side of the barrier if I recall correctly from the Mersen tests. Structure really has very little to do with this. It's a plasma/magnetic/electrical thing. And to my best recollection I recall heavy use of aspirin when I had electromagnetics classes and I recall that at the end of the day, it was pretty much impossible to predict with much certainty exactly what happens in that world...general trends yes but basically the electromagnetic engineers do a lot of lab testing to determine how something will perform in the real world...just too many variables to predict what happens. That was 20 years ago though and the emag software I've seen in the last 10 years is truly amazing in what it can successfully do.