It would be extremely helpful to see the actual XML of your table and table’s children. As you’ve found using indexing in your xpath is useless since that changes with each screen scroll. In my AUT the immediate table rows ( in your example) have no real data. The unique data I need to key from is contained in children inside each row, usually those have a common @resource-id and a unique @text attribute. So I have to queue off that. The annoying thing is the @resource-id is different on the table rows on every page of my app so I can’t easily code a common function to extract/find/etc. table rows.
I’m still trying to crack the case on this myself. You can read more on my adventures here in case there’s some helpful data for you: