You raise some interesting points.
From what I have read, and I may be wrong, but my guess is, that coating many times in fact will provide the failures experienced.
I will try to explain, why I am thinking along these lines.
I have in the past, referred the coatings to an example, of comparing it to a sheet of glass. If your glass sheet is say 1/4 thick, and you try to bend it, it explodes.
However, if you make the glass sheet a few microns thick, then it is very flexible and will deform easily without breaking just like Glass wool.
I am thinking, that by building up several layers of the coating, due to its tough nature, then it may be possible, that the fails mode is reached simply, by the fact of the physical property of the coating becoming much more brittle with the extra thickness.
The smash tests, are certainly well in excess to what is done during sizing, and, no coating comes off, and, these are with two coats baked.
This "theory" seems to be supported by the mere fact, that after passing sizing tests initially, and re-coating again, the smash tests then fail.
Attached is an example of smash tests on coated alloy. The destructive forces used, far exceeds
Attachment 162750 sizing shape changes.
May be, there is room for experimentation to determine at what point the coating will fail/crack, based on thickness applied.