Summary:

When running the following UNNEST query via CBQ when a query memory quota is in place, the query returns partial results and a Request has exceeded memory quota error.

When running the same query with LIMIT n (in this case we used LIMIT 55, though this was also seen with LIMIT 32 in a different cluster) the query runs successfully.

When running the same query again with LIMIT n+1 (we used LIMIT 56), the query returns an error and usedMemory is significantly higher. There seems to be a point n, where usedMemory increases significantly, eg:

No LIMIT clause:

LIMIT n used (55 in this example):

LIMIT n+1 used (56) (usedMemory is significantly higher):

The value of n where the issue starts to occur seems to be different in different clusters.

Steps To Reproduce:

• Spin up a test cluster.

• Deploy the travel-sample bucket.

• On UI under Settings -> Advanced Query Settings -> Memory Quota
  set the limit to 400 MB.

* Execute the following query from CBQ:

* Observe the query returns an error and `usedMemory` is high, with partial results returned:

* Try the same with a high LIMIT clause (I used LIMIT 56) and observe the query also returns the same error:

* Run the query again with lower LIMIT values iteratively, eventually there seems to be a point n where the query runs successfully with no error (this was LIMIT 55 in my cluster):

* Ie. LIMIT 55 works, but LIMIT 56 doesn't.

• Everything below 55 also works, and everything above 56 doesn't.

• This point seems to be different on different clusters - saw the same issue where LIMIT 32 (and below) worked but LIMIT 33 (and higher) showed the issue on a different cluster. 

• Might require testing different values of LIMIT n to reproduce the issue.