Thanks Fons.  Do you have a recommendation for how much headroom to leave 
before the conversion, and if you do clip during the conversion is there
any sort of warning?  I used to leave around 1.5 dB.  The way I used to
go about this was to apply all limiting to the 24-bit 48000 kHz track,
then SR convert, then normalize (to ~ -0.3 dB) or set a final level, and
finally dither/quantize.