About the view performance (view update speed) on the humongous amount of documents

My sincere thanks in advance.

I’m currently working on someone else’s Couchbase system.
And the situations are as below.

We are putting all user purchase records in a bucket “reward”, and which also contains other user information.
Each document is used to check if user get paid or etc…
And we are using view map-reduce feature to make all those documents into report table.
But the report table’s data is slightly lower than the actual value.

The reasons is obvious, we were using view query with default options, and stale option was set to “update_later”,
So some documents were missed by view query.
And the solution is obvious too, set the stale option to “no”.

But, as I indicated above, the data is humongous (around 200 millions and averagely 100 documents income per second),
I was wondering if it’s OK to set stale option to “NO” in this case.
Will there be any problem if I do so?
Or if anyone have ever encountered similar situation to me, how long did it took to retrieve result with stale option “NO” from such an huge bucket?

Of course the best way to find out the result is try it myself,
but since our bucket is in service(And the dev environment bucket is not big enough), I’m feel bit timid to test it directly.

Thank you.