to be on the safe(st) side, i'd rather stick with WAV 32-bit float.
in most if not all cases when recording/writing into "integer" sample file formats, the encoded pcm audio signal often wraps or bounces over the 0dbfs soft-limit, introducing serious, unrecoverable and damaging distortion to the signal--not just the kind of easily flat-top clipping as a regular D/A device might do.
to be on the safe(st) side, i'd rather stick with WAV 32-bit float.
in most if not all cases when recording/writing into "integer" sample file formats, the encoded pcm audio signal often wraps or bounces over the 0dbfs soft-limit, introducing serious, unrecoverable and damaging distortion to the signal--not just the kind of easily flat-top clipping as a regular D/A device might do.
anyway, using a limiter is also an option.
hth.
cheers