Tried that, couldn't see any ripples anymore, just a weird ramp-up in the high treble for the KA17 that didn't match my listening results, though it's mostly happening past 15-ish k, so beyond what I think is my current hearing range. Other than that I could only see that my assumption about the Hiby's filter choice was wrong: it has better, not worse, stopband attenuation, in addition to the better controlled in-band ripples. This ES9281A Pro truly seems to be state-of-the-art DAC tech.
Yep, no difference in ripples, just early/late rolloff for the filters you'd expect to have differences in rolloff. That suggests the problem is outside of the DAC chip, which was already my assumption: if the ripple-like problem is real, it's in the amplification stage.
But I found something else today, my crucial error: in the first listening test I hooked up the KA17's BAL-out to the E/90x electrostatic energizer via its RCA-ins. Because those are 2 separate connectors I was working under the illusion that the channels are staying completely separate, but the E/90x does NOT have balanced topology, its RCAs have the same ground! So I was effectively shorting the BAL-out's L and R returns together, which leads to easily audible scrunched-in soundstage width and recessed presentation along the depth axis, even compared to the KA17's own SE-out! Then when I said "I hear the same via the HE-400i" I was probably basing that on confirmation bias carried over from the estat listening, as today I don't hear that big difference through the 400i, the FC3 and KA17 sound damn near identical.
I need to redo this listening comparison with much shorter switching time and listening time, so I think I will just record some representative song through both devices and put it through the ABX tool. Also need to re-measure the multitone thing and make sure my cable setup isn't shorting the BAL channels' returns (though I'm pretty sure I didn't have that problem in measurements as I could only ever get one channel at a time connected to the line-in anyway, but it's worth re-checking).