diff --git a/src/transport_ofi.h b/src/transport_ofi.h
index fab7cf957..93f4ac6b6 100644
--- a/src/transport_ofi.h
+++ b/src/transport_ofi.h
@@ -697,7 +697,7 @@ void shmem_transport_put_signal_nbi(shmem_transport_ctx_t* ctx, void *target, co
                                       };
 
         do {
-            ret = fi_writemsg(ctx->ep, &msg, FI_DELIVERY_COMPLETE);
+            ret = fi_writemsg(ctx->ep, &msg, FI_DELIVERY_COMPLETE | FI_INJECT);
         } while (try_again(ctx, ret, &polled));
 
         SHMEM_TRANSPORT_OFI_CTX_UNLOCK(ctx);
@@ -785,6 +785,8 @@ void shmem_transport_put_signal_nbi(shmem_transport_ctx_t* ctx, void *target, co
                                          };
 
     do {
+        /* FI_FENCE assures completion of one or more (for fragmentation) prior puts through 
+         * signal delivery */
         ret = fi_atomicmsg(ctx->ep, &msg_signal, FI_DELIVERY_COMPLETE | FI_FENCE | FI_INJECT);
     } while (try_again(ctx, ret, &polled));
 
diff --git a/src/transport_portals4.h b/src/transport_portals4.h
index 4e5e5b435..ed377d2cd 100644
--- a/src/transport_portals4.h
+++ b/src/transport_portals4.h
@@ -647,6 +647,8 @@ static inline
 void shmem_transport_put_signal_nbi(shmem_transport_ctx_t* ctx, void *target, const void *source, size_t len,
                                     uint64_t *sig_addr, uint64_t signal, int pe)
 {
+    /* FIXME: Need to optimize non-blocking put with signal for Portals. Current implementation below keeps  
+     * the "fence" in between data and signal put */
     shmem_transport_put_nbi(ctx, target, source, len, pe);
     shmem_transport_fence(ctx);
     shmem_transport_put_scalar(ctx, sig_addr, &signal, sizeof(uint64_t), pe);