Short answer: not unique 
Long answer: all of the IDs (table, view, column, row, page, line etc) are generated through a regular random upon creation of an object. 10 characters are sufficient to call it a good random. The IDs are created on the client side so potentially there is a little risk of collision but it’s negligible IMO.
However, when you copy a doc, all object IDs are cloned as well into the new doc, not regenerated.
And when you duplicate a table within a doc, all column IDs and row IDs are copied too.
So yes, you have to include Doc ID and Table ID within hashes for your row and column keys.