diff --git a/lct/api/lct.h b/lct/api/lct.h
index d62712b0..96a72a5b 100644
--- a/lct/api/lct.h
+++ b/lct/api/lct.h
@@ -7,6 +7,7 @@ extern "C" {
 
 #include <stdint.h>
 #include <stdio.h>
+#include <climits>
 
 #include "lct_config.h"
 
@@ -24,8 +25,7 @@ LCT_API void LCT_set_rank(int rank);
 LCT_API int LCT_get_rank();
 
 // hostname
-#define LCT_HOSTNAME_MAX_LENGTH 128
-extern char LCT_hostname[LCT_HOSTNAME_MAX_LENGTH];
+extern char LCT_hostname[HOST_NAME_MAX + 1];
 
 // time
 typedef uint64_t LCT_time_t;
diff --git a/lct/lct.cpp b/lct/lct.cpp
index e19a091b..5281286f 100644
--- a/lct/lct.cpp
+++ b/lct/lct.cpp
@@ -4,7 +4,7 @@
 
 LCT_API LCT_log_ctx_t LCT_log_ctx_default = nullptr;
 LCT_API int LCT_rank = -1;
-LCT_API char LCT_hostname[LCT_HOSTNAME_MAX_LENGTH] = "uninitialized";
+LCT_API char LCT_hostname[HOST_NAME_MAX + 1] = "uninitialized";
 
 namespace lct
 {
@@ -17,7 +17,8 @@ void init()
     return;
 
   // initialize hostname
-  gethostname(LCT_hostname, LCT_HOSTNAME_MAX_LENGTH);
+  memset(LCT_hostname, 0, HOST_NAME_MAX + 1);
+  gethostname(LCT_hostname, HOST_NAME_MAX);
 
   // initialize LCT_log_ctx_default
   const char* const log_levels[] = {
diff --git a/lct/pcounter/pcounter.cpp b/lct/pcounter/pcounter.cpp
index 9e4730bd..ab7b186f 100644
--- a/lct/pcounter/pcounter.cpp
+++ b/lct/pcounter/pcounter.cpp
@@ -198,6 +198,7 @@ struct ctx_t {
   explicit ctx_t(const char* name_)
       : dump_ofile(nullptr),
         dump_record_on_the_fly(false),
+        dump_record_on_the_fly_lw(false),
         name(name_),
         id(next_id++),
         record_thread(nullptr),
@@ -218,6 +219,11 @@ struct ctx_t {
         dump_ofilename = "lct_pcounter.%.out";
         dump_record_on_the_fly = true;
         record_interval = 1000000;
+      } else if (mode_str == "on-the-fly-lw") {
+        dump_ofilename = "lct_pcounter.%.out";
+        dump_record_on_the_fly = true;
+        dump_record_on_the_fly_lw = true;
+        record_interval = 1000000;
       }
     }
     // Output file to dump
@@ -257,6 +263,10 @@ struct ctx_t {
     if (getenv("LCT_PCOUNTER_DUMP_ON_THE_FLY")) {
       dump_record_on_the_fly = true;
     }
+    if (getenv("LCT_PCOUNTER_DUMP_ON_THE_FLY_LW")) {
+      dump_record_on_the_fly = true;
+      dump_record_on_the_fly_lw = true;
+    }
     // For now, we just assume there will only be a single thread initializing.
     if (start_time == -1) {
       start_time = LCT_now();
@@ -329,7 +339,7 @@ struct ctx_t {
       bool expected = true;
       if (do_record.compare_exchange_weak(expected, false)) {
         record();
-        if (dump_record_on_the_fly) dump(dump_ofile);
+        if (dump_record_on_the_fly_lw) dump(dump_ofile);
       }
     }
   }
@@ -430,6 +440,7 @@ struct ctx_t {
 
   FILE* dump_ofile;
   bool dump_record_on_the_fly;
+  bool dump_record_on_the_fly_lw;
   std::vector<std::string> counter_names;
   std::vector<std::string> trend_names;
   std::vector<std::string> timer_names;
@@ -452,7 +463,12 @@ struct ctx_t {
 void record_thread_fn(ctx_t* ctx, uint64_t record_interval)
 {
   while (ctx->keep_recording) {
-    ctx->do_record = true;
+    if (ctx->dump_record_on_the_fly_lw)
+      ctx->do_record = true;
+    else if (ctx->dump_record_on_the_fly) {
+      ctx->record();
+      ctx->dump(ctx->dump_ofile);
+    }
     usleep(record_interval);
   }
 }