There are several areas where there's no biasing present for the capacitors, in those areas normal e-cap will fail very quickly due to being reverse biased half the time when sound goes through it. In some areas the capacitors would be biased properly most of the time but at large enough signals they will get reverse biased and while the capacitors can survive it, it will still potentially cause problems in the long term, so better be safe than sorry. It is more about longetivity than measurable sound quality, these capacitors will have much greater lifetime due to lacking self destruct mechanism from reverse biasing and healing mechanism that depends on being forward biased. Distortionwise, the larger the value used the lesser will distortion get at lower frequencies (like 150Hz and below). It can reach 0.1% with under 10µF and more with smaller values. With the 47µF I'm using it stays under 0.01%, but ideally they would be over 100µF but I lack physical space for that (if WB header didn't exist I could use taller capacitors). There are sections about capacitor distortion in Douglas Self's books and there's some resources online also.
Ceramics (unless they are NP0/C0G) types introduce lot of distortion (several % even) due to massive loss of capacitance as the voltage across them increases, for DC blocking you only have the option of using MLCCs that are hopefully X5R type to get best out of a bad situation, X7R are sometimes acceptable too but the other types like Y and Z types are not at all. Even in power supply bypass applications they can be problematic as they lose over 50% of their value at any DC bias (i.e sitting between power rails). They are also microphonic.
Tantalum capacitors cannot handle reverse biasing without pretty much immediate damage and it is not very easy to make sure they're always forward biased, distortion performance is also poor (but I have not been able to find anymeasurement graphs) and they are also microphonic. They don't lose capacitance significantly under bias like most ceramics do and are superior to them on this aspect.
Film capacitors are generally linear, even the ones that distort (polyester types) do so at under 0.001% levels. Polystyrene (fragile and low values) has best performance and so do polypropylene types but they don't come in large values that are not insanely large physically. Polyester types come in largest values but they're still very large (several cm³ for 4.7µF for example). They're also expensive like C0G/NP0 capacitors are but are available in much larger capacitances.
I'm only using NP0/C0G types in the filters and feedback paths.
And as far as SB filtering goes, in the end there's nothing that cuts into audible freq range in this card. Authenticity is not in the consideration at all. If I had an original SB(pro) card the filters gets adjusted as the first thing I do with the card, I don't tolerate subjectively muffled sound 🤣.
