I noticed the srfi text says this about (flexpt x y):

"""
If x is zero, then the result is zero. (C99 pow)
"""

However, C99 says (F.9.4.4 in the n1256 draft):

"""
pow(x, ±0) returns 1 for any x, even a NaN
"""

I see no reason to diverge from C99, so I suspect this is just an
oversight.  We may interpret the SRFI text as delegating the detailed
specification to C99, but in that case, we don't even need to mention
the case where x is zero.  Either mentioning x=y=0 case, or omitting
zero case and let the user look up C99, would clarify the intention.