Sandbox Templates and data selection

Question

If I make a partial copy sandbox and create a sandbox template to go with it, how is the data selected in the case of

Grandparent    
   Parent
       Child

lookup relationships? Is there a guarantee that related records will be selected or does the process that copies the data select the records for each object at random? I'd hope the process would not create orphan records, but the help page does not go into any detail

Sandbox Template Help

Assume that I've selected all three object types in the sandbox template.

I don't have access to an org where I can create a sandbox at present or I'd build a test case.

Edited to add:

I've read elsewhere that there's a hard limit of 10,000 records of any given object - even if you are under the db size limit. Results from a recent partial sandbox creation confirm that limit. Bottom line is that the partial copy sandbox may deliver a lot of orphaned related records.

Answer 1

I believe your main consideration here is the partial sandbox size limitation of 5GB. Data will be copied as faithfully as possible and in consideration of record relationships up until the limit... which leads me to believe it could be the main determinant of whether all the relationships for your selected objects are maintained. There is no way to know if the objects will be copied first or last unfortunately.

Also, if your prod org data is <<5GB, all the data might be copied anyway. So no worries then!