I was just looking for this myself and found some of the answers:
- Which geoprocessing tools honor the parallel processing
I couldn't find a comprehensive list of them other than the ones linked in the other answer, but if you look at the geoprocessing tool reference, you can tell for that tool by the list of Environments it supports near the bottom of the tool's documentation page. If it lists "Parallel Processing Factor" in the environments, then it supports it. Otherwise, it does not. It's not a comprehensive listing, but does let you look up support for a particular tool, at least.
- Are these local or global settings (i.e. can you set the
environment at the beginning of the Arcpy script and all respective
tools will honor the environment setting thereafter?
My understanding is that it acts like other environments and can be both. If you set it on the geoprocessing tool itself when running it, it's local, but if you set it in ArcMap environments or arcpy environments, then it's global to that session. I could be wrong on this, but I haven't read anything indicating it behaves differently.
- Are most geoprocessing tools already set to:
arcpy.env.parallelProcessingFactor = "100%"by default?
No. Not necessarily, at least. It's unspecified. Per this help page for ArcGIS 10.2, the default is to let each tool determine its amount. Given the other documentation on that page indicating optimal settings for each tool based upon whether it's disk-bound or CPU-bound, I'd imagine it can vary between using a single core and creating many times as many threads as CPUs. They don't say that, but that's my interpretation of the default they specify.