diff --git a/src/rulesets/Base/indexing.jl b/src/rulesets/Base/indexing.jl
index 61216bda2..4f082f892 100644
--- a/src/rulesets/Base/indexing.jl
+++ b/src/rulesets/Base/indexing.jl
@@ -262,11 +262,7 @@ end
 # Using Val(dim) here is worth a factor of 2 in this, on Julia 1.8-
 # @btime rrule(eachcol, $([1 2; 3 4]))[2]($([[10, 20], [30, 40]]))
 function ∇eachslice(dys_raw, x::AbstractArray, vd::Val{dim}) where {dim}
-    dys = unthunk(dys_raw)
-    i1 = findfirst(dy -> dy isa AbstractArray, dys)
-    if i1 === nothing  # all slices are Zero!
-        return _zero_fill!(similar(x, float(eltype(x)), axes(x)))
-    end
+    dys = unthunk.(unthunk(dys_raw))
     T = Base.promote_eltype(dys...)
     # The whole point of this gradient is that we can allocate one `dx` array:
     dx = similar(x, T, axes(x))