@manx this will also require some thinking on the libopenmpt side. The BLEP-based resampler is implemented as an interpolation template in the mixer, as that's where it belongs conceptually. However, this is technically not an interpolation with a fixed kernel width that would fit into the current scheme how interpolation is set up in the library interface. Maybe it could also be implement via a ctl instead. Any suggestions are welcome.

I wanted to add some additional API regarding filter selection anyway in order to make it more future-proof with regard to the ability to select filter anti-aliasing on/off. (Which would allow to select between the 8tap kaiser (which does apply crude anti-aliasing lowpass (current default in libopenmpt)) and the 8tap XMMS filter (which does not (currently not available in libopenmpt))).
With regard to anti-aliasing, I was thinking of a ctl, like mixer.interpolation.anti-aliasing=[1|0] (with 1 being default as it's the current behaviour). This would not do anything when applied to filter lengths which do not support it.
For BLEP, I think the setting could be named mixer.interpolation.blep=[1|0] (with 0 being default), which would just override all other interpolation settings in case they do not apply to the BLEP resampler.
The complicated aspect of any such ctl and implementation in general is, how it would interact with per-instrument interpolation filter settings.

I think I would only want to allow the BLEP filter to work with Amiga-based formats anyway - so per-instrument interpolation would not apply. It could be a per-song setting (that can be enabled through a checkbox in OpenMPT and through a ctl in libopenmpt) that effectively replaces the "no interpolation" setting, since that's what it basically is - no interpolation with a reconstrution filter. Another possibility would of course be to make it a choice in the song-specific resampler dropdown list and only show it for those module types.

r8292 adds support for the Amiga resampler.