1) @vector->list and xxxxxx@vector->list both take an optional start and end argument.

Why not @vector->vector and vector->@vector also?

2) Does anything need to be said in the document about how the different floating-point precision vectors interact with Schemes that provide multiple scalar inexact real precisions?

I think it's fairly clear that for values going IN to a f{32,64}vector, the argument must be any inexact real, no matter what the precision, and intuitively if information is lost in the conversion (either by rounding or pegging to +/-INF) that's the users's problem.

But would this be an acceptable result on a system where single-precision is offered while double-precision is the default?

          (f32vector-ref (f32vector 1.5) 0) ==> 1.5f0