When I first ran across this tumblr I was initially intrigued. “What contrapuntal circumstances,” I wondered, “might force such a master of control and planning as Bach to let errors or missteps stand?” But I was disappointed, and the adage “a little knowledge is a dangerous thing” was once again confirmed.
Somewhere in “Free Composition” Schenker observes that parallels in the abstract middleground can be obliterated (to use his term) by diminution in the foreground, and that some foreground parallels can be admitted because they do not represent real voice-leading (ie, one or both of the notes in question are superficial, decorative, and do not comprise real motion – eg, a simple neighbor note). So almost none of these alleged parallel perfect intervals are legitimate parallels. Most often they are the result of a written-out mordent.
And simple proximity alone does not create a cross-relation between a diatonic pitch and a chromatically altered pitch. The cadence in the voices shown in the first example (BWV 248) can be thought of as a distinct event from the movement-initiating diminished 7 (of which the C- natural is a part) which follows in the orchestra, even if in practice they overlap ever-so-slightly. Schenker explained that cross-relations are real only if there is no reason for the diatonic and the chromatically altered pitches to be where they are. If, however, they are fulfilling functions, there is no cross-relation. The vocal tenor C# fulfills the expectation of a cadence in major, set up by the C# in the vocal bass, and the C-natural in the violas, together with D#, heralds harmonic motion to E minor. The other examples of alleged cross-relations are similar.
Only one of the examples on this tumblr represents a legitimate misstep, imo, and it’s not inconceivable that it could be a copyist’s error. It’s the third example – the harmonization of “Du, o schönes Weltgebäude”. I don’t see why Bach wouldn’t have resolved the tenor G# to A, thereby avoiding both the hidden and the overt parallel fifths. Having the full triad at that cadence doesn’t seem like a necessity to me. But like I said, perhaps it’s a copyist’s error.