The only time I remember plainly hearing digital filter ringing was when using a resampler with a very steep filter (unmodified SSRC) and a source or destination sample rate that placed fs/2 well within my hearing range. Neither should be the case in playback of 44.1 kHz or higher material over typical DACs, so that's probably a red herring.
Two more interesting leads to follow up on:
1. Insufficiently suppressed alias components in the ultrasonic range would not only mess with a measurement system, but might also cause audible intermodulation distortion.
2. One thing that tends to change between fast and slow rolloff filters is periodic passband ripple. This is correlated with pre- and post-echoes, and the higher its frequency in the frequency domain, the further apart those are from the main impulse and the less likely they are to be masked audibly (same goes for amplitude). This is primarily an issue for linear phase (FIR) filters, as minimum-phase (IIR) filters do not produce pre-echoes. It is very likely inaudible either way if passband ripple is kept below +/-0.001 dB (perhaps even +/-0.003 dB already), and safely inaudible below +/- 0.0001 dB (= 102 dB down). You can find numerous DACs that are in the green or at least in the yellow in this regard; it's a much greater challenge for ADCs (a lot of the current AKMs will produce about +/-0.06 dB at 48 or 96 kHz).
If you insist, you can always study your DAC's DAC chip datasheet, pick out a setting that results in virtually nil periodic passband ripple, and upsample in software using a known-good resampler (like the SoX resampler plugin for Foobar2000, which even allows you to further reduce ringing by picking an even lower filter cutoff frequency - although the default 95% of fs/2 is plenty fine already - and continuously vary between FIR and IIR filtering). This would essentially eliminate both of the above.
