Perform Scale Tests for all sketch functions across different BQ regions

[nikunjbhartia]

Since there is no good way to automate integration tests at scale at the moment that could potentially have caught some of the recent issues we have seen, can we try the following to ensure we are not missing other potential bugs :

1. Come up with a large enough ( ~ x Billion records ) sample table
2. Create sample queries for each of the sketch functions ( with group by clauses in relevant cases )
3. Execute these queries in each of the BQ regions

Since the sketch functions are deployed in each regional BQ public dataset, performing #3 should be trivial.

[nikunjbhartia]

I am thinking if we can have a script which accepts a BQ project and dataset, and performs all the 3 steps above.
One potential reason why its difficult to reproduce some of the recent failures might be difference in the BQ pricing model
My tests were done on a project with on-demand model, whereas yours would be on BQ editions I assume. Both have different quotas and slot limits which might potentially effect the query execution.