Inconsistencies with multi-argument constructors

Given that Scalar <: SArray, the behavior of
https://github.com/JuliaArrays/StaticArrays.jl/blob/552bc7596c1639d1f66bf781084d3961b36dbd53/src/Scalar.jl#L10
is inconsistent with
https://github.com/JuliaArrays/StaticArrays.jl/blob/552bc7596c1639d1f66bf781084d3961b36dbd53/src/convert.jl#L5

Fortunately, in a few cases there's a very welcome ambiguity error:

julia> sa = Scalar((1,))
Scalar{Int64}((1,))

julia> Scalar(sa)
ERROR: MethodError: Scalar{T} where T(::Scalar{T}) is ambiguous. Candidates:
  (::Type{Scalar{T} where T})(a::AbstractArray{T,0} where T) in StaticArrays at /home/tim/.julia/dev/StaticArrays/src/Scalar.jl:11
  (::Type{Scalar{T} where T})(a::AbstractArray) in StaticArrays at /home/tim/.julia/dev/StaticArrays/src/Scalar.jl:10
  (::Type{SA})(a::StaticArray) where SA<:StaticArray in StaticArrays at /home/tim/.julia/dev/StaticArrays/src/convert.jl:4
Possible fix, define
  Scalar{T} where T(::StaticArray{S,T,0} where T where S<:Tuple)
Stacktrace:
 [1] top-level scope at none:0

but mostly this just silently does things differently for 0 dimensions than for any other dimension.

I don't think you can really have both. Can you choose which behavior you want to keep?

Read the constructor methods in src/Scalar.jl and src/convert.jl at the linked lines, then reproduce the Scalar((1,)) and Scalar(sa) examples in Julia. Resolve which multi-argument and conversion behavior should be retained, with consistent behavior for zero-dimensional and other-dimensional arrays as the completion criterion.