Multiprocessing & in_memory workspace - table writing conflicts?

[HarryBowman]

If I have multiple processes working away, the sample posted at
http://blogs.esri.com/esri/arcgis/2012/09/26/distributed-processing-with-arcgis-%e2%80%93-part-1/

makes it appear that you can write to a single table in_memory\identical  in all the processes.

That would suggest a separate copy of the workspace for each process, which I think it was Bruce Harold says in the comments. But frankly, I am still lost.

What's making me confused is the commentary in this threadhttp://gis.stackexchange.com/questions/35614/concurrent-use-of-in-memory-workspace-in-arcpy

which makes it seem like in_memory is shared across processes and is sort of like a "folder" in that you are OK to write to multiple tables as long as they have different names, but NOT to the same table name. Note, ths gis.stackexchange post is from 2011.

So my question is whether a separate in_memory workspace is created for each process (so all processes end up writing to their own in_memory\table1 or whether it is shared, so you need to change the table names in_memory\table1, in_memory\table2, etc. or something else?

I know I should just code and test, but maybe my test will work but under the hood it is "unsafe."