Seat belts which connect into the restraint and not across a guest are not a safety device, but a measuring instrument. If the seat belt can't reach the buckle attached to the restraint, the restraint is not sufficiently locked. It's used to weed out guest of extra ordinary sizes where their body type would prevent the locking mechanism from being far enough down on the ratchet to safely lock, but otherwise serves no safety purpose. These need to be checked and secured, but if a guest unbuckles it after they have been checked I wouldn't stop a ride over it.
In regard to seat belts that go across the guest, on most modern rides they are supplementary to another restraint such as a lapbar. They are designed to be a secondary restraint and also reduce the distance of travel between the guest and the restraint on rides where the lap bars used an older broader standerd of latches. It's to help with rider comfort where older restraints don't provide a snug enough fit. Ones that may only have 6-10 settings that are 1/2 an inch apart. (If it clicks, it's a legacy system)
Newer restraint systems are infinity adjustable and don't have this problem. In some cases they are also self tightening. Rides built in the last year are now coming standard with electronically monitored restraints, we can tell if it's locked and how much pressure is being pushed on it.
On average it takes 3-6 seconds to properly check a seat system with a supplementary restraint. That number can be lower if they employ the system that allows the operator to tug on the restraint externally from the ride vehicle.
We check both types of restraints, each and every cycle, and we average 3.5 seconds per seat.
I must re-enforce however that what we are observing in terms of through put is so far out of the ordinary that its not even an issue of efficiency. There is no efficiency. We are talking about trains that are stacking for 90+ seconds.