diff --git a/Project.toml b/Project.toml
index 85c8ed0e4..90a297d34 100644
--- a/Project.toml
+++ b/Project.toml
@@ -1,7 +1,7 @@
 name = "KernelAbstractions"
 uuid = "63c18a36-062a-441e-b654-da1e3ab1ce7c"
 authors = ["Valentin Churavy <v.churavy@gmail.com> and contributors"]
-version = "0.9.41"
+version = "0.9.42"
 
 [deps]
 Adapt = "79e6a3ab-5dfb-504d-930d-738a2a938a0e"
diff --git a/src/cpu.jl b/src/cpu.jl
index 7ee786635..f590ff3dd 100644
--- a/src/cpu.jl
+++ b/src/cpu.jl
@@ -111,13 +111,15 @@ function __run(obj, ndrange, iterspace, args, dynamic, static_threads)
     if Nthreads == 1
         __thread_run(1, len, rem, obj, ndrange, iterspace, args, dynamic)
     else
-        if static_threads
-            Threads.@threads :static for tid in 1:Nthreads
-                __thread_run(tid, len, rem, obj, ndrange, iterspace, args, dynamic)
-            end
-        else
-            @sync for tid in 1:Nthreads
-                Threads.@spawn __thread_run(tid, len, rem, obj, ndrange, iterspace, args, dynamic)
+        let len = len, rem = rem, Nthreads = Nthreads
+            if static_threads
+                Threads.@threads :static for tid in 1:Nthreads
+                    __thread_run(tid, len, rem, obj, ndrange, iterspace, args, dynamic)
+                end
+            else
+                @sync for tid in 1:Nthreads
+                    Threads.@spawn __thread_run(tid, len, rem, obj, ndrange, iterspace, args, dynamic)
+                end
             end
         end
     end
diff --git a/test/runtests.jl b/test/runtests.jl
index f395b70dc..da102bbab 100644
--- a/test/runtests.jl
+++ b/test/runtests.jl
@@ -17,6 +17,11 @@ A = zeros(Int, Threads.nthreads())
 kern_static(CPU(static = true), (1,))(A, ndrange = length(A))
 @test A == 1:Threads.nthreads()
 
+@testset "__run does not box (no per-launch Core.Box)" begin
+    ci = only(code_lowered(KernelAbstractions.__run, NTuple{6, Any}))
+    @test !any(e -> occursin("Box", string(e)), ci.code)
+end
+
 @kernel cpu = false function my_no_cpu_kernel(a)
 end
 @test_throws ErrorException("This kernel is unavailable for backend CPU") my_no_cpu_kernel(CPU())