Minor Description
The implementation of Duration operator*=() is as follows:
Duration& operator*=(double multiplier)
{
nanos = static_cast<int64_t>(nanos * multiplier);
return *this;
}
A similar pattern is implemented for all the operators. This means that, even when multiplying by int64_t (underlying type of nanos), we lose precision.
While Review #64729 removes the conversion warnings from int and size_t to double, it purposefully does not address fixing the precision of these operators (as that'll be a change in behavior, albeit slight, and should be done for the whole class at once).