It would seem that when case marking in sluicing is not directly linked to case marking in nonsluiced associates, it involves default case, as is plausibly the case in the Turkish data cited above.
To return to sluicing, I note that Dutch has a construction with van 'of, sometimes referred to as quotative van, which may involve a kind of non-wh sluicing (Temmerman 2013 calls them embedded fragment answers).
Just like in the case of sluicing, we have a remnant instead of a full embedded clause.
The difference between (11a) and (11c) suggests that pied-piping of prepositions is obligatory in sluicing. However, (11b) is at odds with that suggestion.
This suggests very strongly that sluicing is not fed by wh-movement, even though much of the literature suggests otherwise, since wh-movement in Dutch does not allow for preposition stranding, except when the wh-word is also an R-word: waar.
This is to be expected, if embedded fragment answers are a special kind of sluicing. It would be interesting to do a corpus study of pied-piping (or lack thereof) in sluicing (see Nykiel 2013 for a study of preposition omission in various ellipsis constructions in English, including sluicing, short answers and so-called 'split questions').
Clearly, OE already witnesses integration of the syntactic side of sluicing with a non-syntactic one inasmuch as syntactic reconstruction is not a viable option in passages like (16)-(18), though it could be posited for (15) and (19)-(20).
To be honest, initially it was my intuition that sluicing, at some point, had only been constrained by syntactic effects, with sprouting arising later on and gaining ground step by step.
While it is essential that one appeal to the authentic/translated distinction, the distinction itself fails to shed new light on how sluicing and sprouting came about.
To stretch this point to its logical conclusion, an approach along these lines could also explain and interpret inference-based and situationally-controlled sluicing, recorded in PDE, as driven by mental images/notes evoked by overt antecedents.
Overall, it is essential that linguists not lose sight of syntax, since the DOE corpus sees sluicing controlled by syntax in 142 (entire or partial syntactic control) out of 165 instances.