diff --git a/core/partitions/src/messages_writer.rs b/core/partitions/src/messages_writer.rs
index 9e61ec1bef..d689127190 100644
--- a/core/partitions/src/messages_writer.rs
+++ b/core/partitions/src/messages_writer.rs
@@ -178,9 +178,9 @@ fn preallocate_file(file: &File, file_path: &str, len: u64) {
     // sets `thread_pool_limit(0)` on the shard proactor, so `spawn_blocking`
     // has no worker to park a task on and compio panics the shard outright with
     // "the thread pool is needed but no worker thread is running". (That limit
-    // is skipped on macOS/aarch64, where the pool does exist -- see the FIXME
-    // there -- so the panic is Linux-and-most-targets, not universal. This arm
-    // is Linux-only regardless.)
+    // is skipped on macOS, whose polling driver routes fs through the pool, so
+    // the panic is Linux-and-most-targets, not universal. This arm is
+    // Linux-only regardless.)
     //
     // The cost is acceptable only because of what this call is: a metadata-only
     // extent reservation, microseconds on the local filesystems this option
diff --git a/core/server_common/src/executor.rs b/core/server_common/src/executor.rs
index 8fb8cc8422..8ad8f9920f 100644
--- a/core/server_common/src/executor.rs
+++ b/core/server_common/src/executor.rs
@@ -84,7 +84,7 @@ pub fn create_shard_executor() -> Result<Runtime, std::io::Error> {
     // io_uring targets keep the zero limit: no blocking pool exists on shard
     // threads, which `core/partitions` messages_writer relies on to justify
     // running fallocate inline (`spawn_blocking` would hit that same panic).
-    #[cfg(not(all(target_os = "macos", target_arch = "aarch64")))]
+    #[cfg(not(target_os = "macos"))]
     proactor.thread_pool_limit(0);
 
     compio::runtime::RuntimeBuilder::new()