Well, actually I based what I said on having seen the outcome of front impact tests on many early Minis, which were used because they were cheap. They were actually testing the crash test dummies at the time, not the cars, but as I was on Mini number 2 at the time, I paid close attention to how they crumpled, and in no case did the engine come through the bulkhead, but almost every time the subframe, all solid mounted in those days, transferred the impact to the floor, which buckled badly, trapping, and sometimes breaking, the feet of the dummy. Upper body sensors usually said that the dummies were dead or seriously injured, due to high seatbelt loads, and over 50mm chest compression. The engine in fact helped to spread the load across the bulkhead. The engine is initially travelling at the same speed as the Mini, and its inertia will cause it to try to continue travelling forwards. When it meets something solid enough to stop it, then, and only then, is it in any danger of penetrating the bulkhead, which does of course happen in severe crashes, but these are not survivable anyway due to the rapid deceleration of the chest and upper body. Airbags do not help, only a longer crumple zone to lower the peak deceleration. A very much stronger bulkhead does not help either, for the same reason, peak deceleration. But making the subframe side rails crumple will help.
My particular interest was the sensors on the side impact dummy, and it was found that it is possible to make a reasonably thin door that is sufficiently strong to resist penetration by an incoming stupidly heavy car (Chelsea Tractors had not been invented then), but it would only kill the occupants due to high lateral acceleration imparted to the whole car. So along comes the Smart car, actually a very dumb car indeed to own for many reasons, which purports to protect its occupants by having a very rigid passenger shell. Well, I have news for all Dumb car owners, it doesn't work in most crashes. It relies ENTIRELY on activating the crumple zones of the other vehicle, so it is intentionally designed to kill YOU in your Mini or other relatively fragile car, well in keeping with the general ethos of Mercedes. It DOES NOT WORK AT ALL in a serious collision with something unyielding such as some Chelsea Tractors, all HGVs, brick walls and many other things. Hit any of those at speed in a Dumb car, and you will be found strapped in your seat, airbag burns on your face, and well and truly DEAD due to the unsurvivable deceleration imposed by the rigid shell.
You can't look at a structure and see how it will behave in any particular accident, only in general terms, as the initial buckling will be different every time, so even computer simulations are not actually very good. But you can look at something and identify the main load paths, although calculating the magnitudes of the loads in complex conditions is always challenging, even to an expert with computer assistance.
The early Honda Civic had a severe problem meeting the initial front impact test of the 1980s, and that was a straight impact, not offset, so it must have been a very poor car. Anyway, the regulations at the time only required (and probably still do) one pass, so they threw 160 cars at the test barrier until, finally, one initially crumpled in a way that allowed it to scrape a pass. The car was then type approved and went on sale, but I don't believe that anyone at TRRL ever bought one.
Modern Minis seem to do much better in crashes, despite there being virtually no change to the body structure. We have seen the results of some crashes here on the forum where a solid mounted front subframe may well have resulted in trapped feet. The woefully inadequate toeboard mount rubbers break regularly at the best of times, as they are not safety through-bolted, an amazing omission for a structural mount, unless it is for the very good reason that they are supposed to break in an accident, allowing the subframe legs to slide back under the floor. Manufacturers often sneak in a mod for a plausible reason other than the real one, e.g. we all have heard that the Mini engine was turned around to cure carburettor icing, and that first appeared in Pomeroy's book, early on, which I have just been reading. But, we now know that the reason was actually that the inertia of two large drop gears was much too great for the poor old synchro cones in the A35 geartrain, and as inertia of a solid disk varies as the 4th power of its diameter (ok, the gears need not be solid discs, part can be machined away, but we are being approximate), the inertia, and thus the stored kinetic energy, of 3 small gears, some of which must be dumped as heat in the synchros, is far less than that of 2 big ones. So, were the rubber mounts introduced to make the car quieter, when other methods of doing that, such as improved insulation materials are available, or was it actually sneaked in to modify the behaviour in a crash, something that car manufacturers don't like to talk about publicly? Remember that crash tests were being introduced at about that time. Did it make the car quieter? In my opinion, no, but I did not have equipment available to measure it accurately, and in any case changing from A to A+ engine and especially gearbox made significant differences.
I would still convert a rubber mounted car to solid, to make it handle properly, and was always intending that the stiffening plate would extend to at least the tunnel on one side and the inner wing on the other, with a short extension rearwards on the floor to stiffen it in a fore and aft direction, to attempt to minimise the floor crumpling in the event of a hard frontal impact. A roll cage is very inappropriate in a road car and only increases accident risk (higher centre of mass, and it encourages idiotic boy-racer driving), and insurance costs, but I do know that some of them have a tubular member in that area to prevent subframe intrusion.
