As footnote, the error in survival:::untangle.specials is that it assumed that if
attr(myterms, 'specials')[['strata']] was = to 4, then one could use myterms[-4] to remove
the strata term. Not so. In the model y ~ x1 + x2 + strata(x3) the attritube will
be 4 -- the response counts --- but to remove it I need to use [-3] since the response
does not count in [.terms or drop.terms.
Is this an inconsistency that should be documented and/or repaired? What would break if
we did? I don't know. By chance, all the usages in the survival package happened after
a call to delete.response so it would be immune to such a change.