diff --git a/CMakeLists.txt b/CMakeLists.txt
index 73baecc34..8dc8bd447 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -21,7 +21,6 @@ else()
   set(ENABLE_ACC_CHECK OFF)
 endif()
 
-option(CELERITY_USE_MIMALLOC "Use the mimalloc memory allocator" ON)
 option(CELERITY_ACCESSOR_BOUNDARY_CHECK "Enable accessor boundary check" ${ENABLE_ACC_CHECK})
 
 if(CELERITY_ACCESSOR_BOUNDARY_CHECK)
@@ -119,6 +118,15 @@ if(NOT CMAKE_BUILD_TYPE)
   set(CMAKE_BUILD_TYPE "Release" CACHE STRING "CMake Build Type" FORCE)
 endif()
 
+if(CELERITY_SYCL_IMPL STREQUAL "DPC++")
+  # See https://github.com/oneapi-src/unified-runtime/issues/803
+  message(STATUS "Not enabling mimalloc by default because it breaks with oneAPI plugin loading")
+  set(CELERITY_USE_MIMALLOC_DEFAULT OFF)
+else()
+  set(CELERITY_USE_MIMALLOC_DEFAULT ON)
+endif()
+option(CELERITY_USE_MIMALLOC "Use the mimalloc memory allocator" ${CELERITY_USE_MIMALLOC_DEFAULT})
+
 # 3rdparty dependencies
 include(FetchContent)
 
diff --git a/cmake/celerity-config.cmake.in b/cmake/celerity-config.cmake.in
index d8d784f19..9d7f6d9e1 100644
--- a/cmake/celerity-config.cmake.in
+++ b/cmake/celerity-config.cmake.in
@@ -23,7 +23,9 @@ find_dependency(fmt REQUIRED)
 find_dependency(spdlog REQUIRED)
 find_dependency(small_vector REQUIRED)
 find_dependency(libenvpp REQUIRED)
-find_dependency(mimalloc REQUIRED)
+if(@CELERITY_USE_MIMALLOC@)
+  find_dependency(mimalloc REQUIRED)
+endif()
 if(@CELERITY_ENABLE_CUDA_BACKEND@)
   if(CMAKE_VERSION VERSION_GREATER_EQUAL "3.17")
     find_dependency(CUDAToolkit REQUIRED)
diff --git a/src/worker_job.cc b/src/worker_job.cc
index c30295ae9..9ef1abcd2 100644
--- a/src/worker_job.cc
+++ b/src/worker_job.cc
@@ -217,7 +217,7 @@ namespace detail {
 				try {
 					const auto info = m_buffer_mngr.access_device_buffer(bid, mode, sr);
 #if CELERITY_ACCESSOR_BOUNDARY_CHECK
-					auto* const oob_idx = sycl::malloc_shared<id<3>>(2, m_queue.get_sycl_queue());
+					auto* const oob_idx = sycl::malloc_host<id<3>>(2, m_queue.get_sycl_queue());
 					assert(oob_idx != nullptr);
 					constexpr size_t size_t_max = std::numeric_limits<size_t>::max();
 					const auto buffer_dims = m_buffer_mngr.get_buffer_info(bid).dimensions;
diff --git a/test/backend_tests.cc b/test/backend_tests.cc
index 902801ea1..d6d899410 100644
--- a/test/backend_tests.cc
+++ b/test/backend_tests.cc
@@ -124,7 +124,14 @@ TEMPLATE_TEST_CASE_SIG("memcpy_strided_device allows to copy between the same de
 	const size_t platform_id = GENERATE(Catch::Generators::range(size_t(0), sycl::platform::get_platforms().size()));
 	const auto test_type = GENERATE(copy_test_type::intra_device, copy_test_type::inter_device, copy_test_type::host_to_device, copy_test_type::device_to_host);
 	CAPTURE(platform_id, test_type);
+
 	const auto platform = sycl::platform::get_platforms()[platform_id];
+	CAPTURE(platform.get_info<sycl::info::platform::name>());
+
+	if(platform.get_info<sycl::info::platform::name>().substr(0, 5) == "Intel" && test_type == copy_test_type::inter_device) {
+		SKIP("Inter-GPU copy appears to currently be broken on Intel OpenCL / Level Zero");
+	}
+
 	auto [src, tgt] = ([&] {
 		try {
 			return select_source_and_target(test_type, platform);