You guys realize that, as we are talking about all of this, part of the reason we can differ in our opinions is that there are NO GOOD DATA. There simply aren't enough people who arrest underwater -- where somebody else sees it and can help -- to generate a database that will help us decide what the best course of action is. All the stuff we're taught is somebody trying to think the problem through ON LAND and come up with a good solution.
On land, the best intervention is shock. That's known. On land, the most common reason for a cardiac arrest is coronary artery disease, and many times, the rhythm is Vtach or Vfib, which are amenable to shock. Getting the person out of the lethal rhythm is the BEST intervention. CPR is critical until the shock is available, and between shocks, but shock is best.
In the water, it's more likely that the arrest is respiratory . . . and primary respiratory arrest (eg. in this case, drowning) tends to lead to asystole, or no heart rhythm. You can't shock somebody out of asystole. Rarely, if you correct the condition that created the asystole and do enough effective CPR, you might get a shockable rhythm. Generally, if asystole has persisted more than a short time, we call the code.
Heart disease is still the number one killer in the US, and a large number of people arrest out of the hospital. We collect data on these arrests, and the interventions, and despite doing that for decades, the American Heart Association KEEPS changing the ACLS protocols, as we discover the things we were taught to do for cardiac arrest patients really don't work very well. When you consider the small number of arrested divers in comparison, it is easy to see that nobody really knows what we should do.
If you don't have medical training, do what your Rescue class taught you. If you do have medical training, apply what you know and your common sense to the situation, and do what you think is reasonable and appropriate. Don't create a second victim. Don't take extreme risks in the setting of an unlikely good outcome.