Error when computing derivative of function involving a complex matrix

For example consider the following minimal program:

function f4(x)
abs( (Complex[1] * x)[1] )
end

tp = InstructionTape()
at = ReverseDiff.TrackedReal(0.41992711708322633, 0, tp)
ot = f4(at)
ReverseDiff.seed!(ot)
ReverseDiff.reverse_pass!(tp)
println(deriv(at))

TypeError: in TrackedReal, in V, expected V<:Real, got Type{Complex{Float64}}

Stacktrace:
[1] ReverseDiff.TrackedArray(::Array{Complex{Float64},1}, ::Array{Int64,1}, ::Array{ReverseDiff.AbstractInstruction,1}) at /Users/wmoses/.julia/packages/ReverseDiff/qmgw8/src/tracked.jl:86
[2] track(::Array{Complex{Float64},1}, ::Type{Int64}, ::Array{ReverseDiff.AbstractInstruction,1}) at /Users/wmoses/.julia/packages/ReverseDiff/qmgw8/src/tracked.jl:387
[3] broadcast_mul(::Array{Complex,1}, ::TrackedReal{Float64,Int64,Nothing}, ::Type{Int64}) at /Users/wmoses/.julia/packages/ReverseDiff/qmgw8/src/derivatives/elementwise.jl:421
[4] broadcast at /Users/wmoses/.julia/packages/ReverseDiff/qmgw8/src/derivatives/elementwise.jl:343 [inlined]
[5] * at ./arraymath.jl:55 [inlined]
[6] f4(::TrackedReal{Float64,Int64,Nothing}) at ./In[165]:2
[7] top-level scope at In[165]:7

On the other hand the following succeeds:

function f4(x)
abs( (Complex[x])[1] )
end

tp = InstructionTape()
at = ReverseDiff.TrackedReal(0.41992711708322633, 0, tp)
ot = f4(at)
ReverseDiff.seed!(ot)
ReverseDiff.reverse_pass!(tp)
println(deriv(at))

Start with the minimal reproducer in the issue and inspect the TrackedArray handling in src/tracked.jl, then follow the broadcast_mul path in src/derivatives/elementwise.jl. Compare the failing Complex[1] * x case with the working Complex[x] case; done means the failing example completes without the TypeError and reports its derivative.