I've got a general question about what are the recommended use cases when deciding to use in_memory vs scratch workspace/folder/GDB for intermediate (i.e. temp) data in a Python script. I've searched and the only real answers I have seen indicate that a scratch GDB will let you take advantage of more GDB features and that you have to explicitly delete your temp data in the scratch GDB.
So, when would I want my script to use in_memory vs scratch workspace? Is there a difference in terms of performance and memory resources? Are there different recommendations when dealing with Desktop, Pro, and Server?
In ArcGIS Pro 2.3, in_memory has been changed and improved. The online help isn't available until the release is made out … very soon... but I would
Currently the help shows...
and currently some tools don't support it
Check those links when 2.3 is released