I was wondering to what extent the agreement pattern exemplified with the following sentences drawn from Cicero's De Amicitia can be regarded as the most natural one. I'm asking this question since, for example, in my native language (Catalan), one would expect singular agreement here (cf. constituendum est), the indirect interrogative subordinate clause being its neuter singular subject. However, it seems that the attraction pattern exemplified below could be claimed to be the most natural one here. Is this true?
Constituendi autem sunt qui sint in amicitia fines et quasi termini diligendi (Cic. Amic. 56)
Quidam saepe in parva pecunia perspiciuntur quam sint leves; quidam, quos parva movere non potuit, cognoscuntur in magna. (Cic. Amic. 63).
Ben Kovitz's example Perspicio te, Miles Gloriose, quam levis sis, which involves an instantiation of the so-called proleptic accusative, is really illuminating. This construction could well be at the basis of the passive construction of the example above Quidam saepe in parva pecunia perspiciuntur quam sint leves: i.e., its nominative subject quidam could correspond to a subjacent proleptic accusative object in the active voice. Cf. also these interesting examples from Cicero & Terentius, respectively: nosti uirum quam tectus (Cic. Att. 14.21.2); patrem nouisti ad has res quam sit perspicax (Ter. Haut. 370).
Informally speaking, it seems that the two sentences above involve a sort of blending of two different syntactic structures:
(1) Constituendi sunt fines & (2) constituendum est qui sint in amicitia fines;
(1) Quidam saepe in parva pecunia perspiciuntur & (2) Perspicitur quam (quidam) sint leves.
Is there any other alternative to this blending? (see Ben Kovitz's skeptical comment below: there is "no need for unusual blending").
Cf. also the following parallelisms:
Constituendi autem sunt (...) in amicitia fines (Cic. Amic. 56) // Alius igitur finis verae amicitiae constituendus est (Cic. Amic. 59).
Quidam saepe in parva pecunia perspiciuntur (...) // quidam (...) cognoscuntur in magna. (Cic. Amic. 63).