diff --git a/services/debugLog.js b/services/debugLog.js
index fa75f38..0150fe6 100644
--- a/services/debugLog.js
+++ b/services/debugLog.js
@@ -11,6 +11,24 @@ function setClient(c) {
   client = c;
 }
 
+// --- PII redaction ---
+
+// Email addresses (loose regex — covers most RFC 5321 local parts that show up
+// in support traffic) and Discord snowflakes (18–20 digit numeric IDs) get
+// redacted before stack/message text reaches the debug channel. Both can land
+// in error stacks via senderEmail interpolation, channel IDs in error
+// messages, etc. — redacting at the boundary keeps the debug channel useful
+// for triage without leaking customer addresses or staff member IDs.
+const EMAIL_REDACT_RE = /[\w.+-]+@[\w.-]+\.\w+/g;
+const SNOWFLAKE_REDACT_RE = /\b\d{18,20}\b/g;
+
+function redactPII(s) {
+  if (s == null) return '';
+  return String(s)
+    .replace(EMAIL_REDACT_RE, '[EMAIL_REDACTED]')
+    .replace(SNOWFLAKE_REDACT_RE, '[ID_REDACTED]');
+}
+
 // --- Helpers ---
 
 async function sendToChannel(channelId, embed, overrideClient) {
@@ -38,9 +56,10 @@ async function logError(context, error, interaction = null, overrideClient = nul
     const commandLine = (interaction?.commandName || interaction?.customId)
       ? `Command/Button: ${interaction.commandName || interaction.customId}\n`
       : '';
-    const stack = (error.stack || error.message || String(error)).slice(0, 1500);
+    const message = redactPII(error.message || String(error));
+    const stack = redactPII(error.stack || error.message || String(error)).slice(0, 1500);
     await channel.send({
-      content: `\`[${context}]\` ${error.message || String(error)}\n${userLine}${commandLine}\n\`\`\`${stack}\`\`\``
+      content: `\`[${context}]\` ${message}\n${userLine}${commandLine}\n\`\`\`${stack}\`\`\``
     });
   } catch (_) {
     // ignore send failures
@@ -52,7 +71,7 @@ async function logError(context, error, interaction = null, overrideClient = nul
 async function logWarn(context, message, overrideClient = null) {
   const embed = new EmbedBuilder()
     .setTitle(`Warning: ${context}`)
-    .setDescription(String(message).slice(0, 4000))
+    .setDescription(redactPII(String(message)).slice(0, 4000))
     .setColor(0xFFFF00)
     .setTimestamp();
   await sendToChannel(CONFIG.DEBUGGING_CHANNEL_ID, embed, overrideClient);
diff --git a/services/staffThread.js b/services/staffThread.js
index 52ed213..cd436da 100644
--- a/services/staffThread.js
+++ b/services/staffThread.js
@@ -11,9 +11,10 @@
  *   is logged via logWarn.
  * - invitable: false means only staff with MANAGE_THREADS can add additional
  *   members — this is intentional for privacy.
- * - guild.members.fetch() in addRoleMembersToThread can be slow on large
- *   servers. The 300ms delay between adds avoids the thread member add rate
- *   limit (approximately 5/second).
+ * - addRoleMembersToThread reads from role.members (cache-derived) and only
+ *   falls back to a scoped guild.members.fetch on cache miss. The 300ms
+ *   delay between adds avoids the thread member add rate limit (~5/sec).
+ *   It runs via setImmediate so it doesn't block ticket creation.
  */
 const { ChannelType } = require('discord.js');
 const { CONFIG } = require('../config');
@@ -39,7 +40,11 @@ async function createStaffThread(channel, client) {
     });
 
     if (CONFIG.STAFF_THREAD_AUTO_ADD_ROLE && CONFIG.STAFF_THREAD_ROLE_ID) {
-      await addRoleMembersToThread(thread, channel.guild, client);
+      // Run off the critical path — the add loop is rate-limited at 300ms per
+      // member and would block ticket creation for ~15s on a 50-member role.
+      setImmediate(() => {
+        addRoleMembersToThread(thread, channel.guild, client).catch(() => {});
+      });
     }
 
     return thread;
@@ -55,16 +60,26 @@ async function createStaffThread(channel, client) {
 
 /**
  * Add all members of the staff role to the thread.
+ *
+ * Prefers role.members (computed from guild.members.cache, kept in sync via
+ * the GuildMembers gateway intent — see broccolini-discord.js intents). Only
+ * falls back to a scoped guild.members.fetch on cache miss (e.g. cold cache
+ * just after restart). Previously called the unscoped guild.members.fetch()
+ * on every ticket creation, which chunked all members of the guild — wasted
+ * gateway/REST budget and added ~15s to ticket creation on busy guilds.
  */
 async function addRoleMembersToThread(thread, guild, client) {
   try {
     const role = await guild.roles.fetch(CONFIG.STAFF_THREAD_ROLE_ID).catch(() => null);
     if (!role) return;
 
-    await guild.members.fetch();
-    const members = guild.members.cache.filter(m =>
-      m.roles.cache.has(CONFIG.STAFF_THREAD_ROLE_ID) && !m.user.bot
-    );
+    let members = role.members.filter(m => !m.user.bot);
+    if (members.size === 0) {
+      // Cache cold (first ticket after restart). withPresences: false skips
+      // the presence sync, which is irrelevant for thread-add and expensive.
+      await guild.members.fetch({ withPresences: false }).catch(() => {});
+      members = role.members.filter(m => !m.user.bot);
+    }
 
     for (const [, member] of members) {
       await thread.members.add(member.id).catch(() => {});