As for now it seems, that notablescan is defined at server level only. I would like this to be a parameter that could be defined at a lower level e.g. in the connection-string.
This would allow one to ensure, that all queries done from production code uses indexes and still I will be able to do ad-hoc queries on my production data in debugging scenarios.
This feature may naturally be overruled if notablescan is ON at server level.
In case notablescan could be defined in the query string, the only change needed to the documentation is instruction hereof.