diff --git a/src/network/Runtime.zig b/src/network/Runtime.zig
index b1b606d8..8de17da1 100644
--- a/src/network/Runtime.zig
+++ b/src/network/Runtime.zig
@@ -60,7 +60,9 @@ wakeup_pipe: [2]posix.fd_t = .{ -1, -1 },
 
 shutdown: std.atomic.Value(bool) = .init(false),
 
-// Async HTTP requests (e.g. telemetry). Created on demand.
+// Multi is a heavy structure that can consume up to 2MB of RAM.
+// Currently, Runtime is used sparingly, and we only create it on demand.
+// When Runtime becomes truly shared, it should become a regular field.
 multi: ?*libcurl.CurlM = null,
 submission_mutex: std.Thread.Mutex = .{},
 submission_queue: std.DoublyLinkedList = .{},
@@ -304,6 +306,13 @@ pub fn run(self: *Runtime) void {
     var drain_buf: [64]u8 = undefined;
     var running_handles: c_int = 0;
 
+    const poll_fd = &self.pollfds[0];
+    const listen_fd = &self.pollfds[1];
+
+    // Please note that receiving a shutdown command does not terminate all connections.
+    // When gracefully shutting down a server, we at least want to send the remaining
+    // telemetry, but we stop accepting new connections. It is the responsibility
+    // of external code to terminate its requests upon shutdown.
     while (true) {
         self.drainQueue();
 
@@ -325,15 +334,15 @@ pub fn run(self: *Runtime) void {
         };
 
         // check wakeup pipe
-        if (self.pollfds[0].revents != 0) {
-            self.pollfds[0].revents = 0;
+        if (poll_fd.revents != 0) {
+            poll_fd.revents = 0;
             while (true)
                 _ = posix.read(self.wakeup_pipe[0], &drain_buf) catch break;
         }
 
         // accept new connections
-        if (self.pollfds[1].revents != 0) {
-            self.pollfds[1].revents = 0;
+        if (listen_fd.revents != 0) {
+            listen_fd.revents = 0;
             self.acceptConnections();
         }
 
diff --git a/src/telemetry/lightpanda.zig b/src/telemetry/lightpanda.zig
index 1e723ee8..ee831ddb 100644
--- a/src/telemetry/lightpanda.zig
+++ b/src/telemetry/lightpanda.zig
@@ -12,7 +12,8 @@ const Runtime = @import("../network/Runtime.zig");
 const Connection = @import("../network/http.zig").Connection;
 
 const URL = "https://telemetry.lightpanda.io";
-const MAX_BATCH_SIZE = 20;
+const BATCH_SIZE = 20;
+const BUFFER_SIZE = BATCH_SIZE * 2;
 
 const LightPanda = @This();
 
@@ -21,7 +22,7 @@ runtime: *Runtime,
 mutex: std.Thread.Mutex = .{},
 
 pcount: usize = 0,
-pending: [MAX_BATCH_SIZE * 2]LightPandaEvent = undefined,
+pending: [BUFFER_SIZE]LightPandaEvent = undefined,
 
 pub fn init(app: *App) !LightPanda {
     return .{
@@ -39,6 +40,11 @@ pub fn send(self: *LightPanda, iid: ?[]const u8, run_mode: Config.RunMode, raw_e
         self.mutex.lock();
         defer self.mutex.unlock();
 
+        if (self.pcount == BUFFER_SIZE) {
+            log.err(.telemetry, "telemetry buffer exhausted", .{});
+            return;
+        }
+
         self.pending[self.pcount] = .{
             .iid = iid,
             .mode = run_mode,
@@ -49,7 +55,7 @@ pub fn send(self: *LightPanda, iid: ?[]const u8, run_mode: Config.RunMode, raw_e
         break :blk self.pcount;
     };
 
-    if (pending_count >= MAX_BATCH_SIZE) {
+    if (pending_count >= BATCH_SIZE) {
         self.flush();
     }
 }
@@ -68,8 +74,6 @@ fn postEvent(self: *LightPanda) !void {
     defer self.mutex.unlock();
 
     const events = self.pending[0..self.pcount];
-    self.pcount = 0;
-
     if (events.len == 0) return;
 
     for (events) |*event| {
@@ -84,6 +88,7 @@ fn postEvent(self: *LightPanda) !void {
     try conn.setMethod(.POST);
     try conn.setBody(writer.written());
 
+    self.pcount = 0;
     self.runtime.submitRequest(conn);
 }