IMO we tend to overthink the frame/body dilemma. It is true that the body could theoretically detach from the chassis in an accident and it is true that there are period photos of this occurring. However, consider that those photos probably overrepresent the frequency of that type of accident. It's a very dramatic image!
But we have no data on how often that really happened at the time or how often it happens now. I have not heard of any Model A body-detachment accidents in the recent past. The vast majority of accidents I've heard of involve:
- doors open without warning
- driver loses control and hits curb/tree/ditch
- low speed collision at intersection
In these scenarios the most critical objectives are to keep the occupants inside the cabin and prevent them from hitting unyielding elements of the car like the body supports or steering column. This is true even for open-car designs, as the accident investigator commenter on your video showed.
To achieve both these objectives, you really need a shoulder belt, but that's not always possible, so let's set that aside. To optimize the restraining function of a lap belt, the belt needs to keep as much of the occupants' mass as possible from moving forward within the cabin. This is quite hard, as you'll see if you watch any of those slo-mo videos of crash test dummies. Our goal is to reduce the severity of any injury by preventing the center of the body from hitting anything.
The force on the belt at the instant of impact can be several times the weight of the occupant. If you have the average 60-year-old married couple in a bench seat, that weight is 372 lbs. Add to that the weight of the bench seat itself. So the reinforcing steel that you install needs to be sufficiently robust to deform minimally or not at all under, let's say, 1500 lbs of force, which might be applied off-axis. If it deforms, then the occupant will stay inside the cabin but more of their mass will be moving forward or sideways, causing greater injury when they hit the firewall or the steering column.
So now here is where I differ from your method. I think we should recommend that people attach the seat belts to the chassis because:
- Most owners are not engineers.
- Owners will tend to underestimate the amount of reinforcement required to successfully restrain them in an impact.
- Reinforcing via the chassis is much more likely to provide sufficient reinforcement.
- The risk of this backfiring because the accident is so severe that the body detaches is incredibly small, much smaller than the risk that the owner under-designs the reinforcement, leading to the belt attachment point deforming and the occupant being more severely injured through impact with the windshield, door, etc.