[pwdisswordfish2]

That phrasing seemingly implies that there was a generally-agreed intended behaviour for that case that was unfortunately neither documented nor actually exhibited. But that was not the case. The overflow behaviour of float-to-int casts was undefined, full stop. There was some consensus as to what such casts should not do, but no agreement on what the actual behaviour should be. Eventually, the discussion settled on defining casting overflow to saturate, and the implementation was altered to match.