fix(s3): keep-alive agents + long timeouts to end socket starvation

Root cause of stuck 'processing', failed deletes, and dead playback:

The mam-api proxies media (/video, /hls pipe the full S3 body through
Express), holding long-lived streaming sockets. With the SDK's default
http agents (no keep-alive, unbounded but unpooled) those streams starved
control-plane calls — DeleteObject and the proxy worker's master download
— which timed out (10s connectionTimeout) in bursts.

Fixes:
- mam-api S3 client: dedicated keep-alive http/https Agents (maxSockets 256)
  + requestTimeout raised 30s→300s so large master GETs finish.
- worker S3 client: previously had NO handler config at all (SDK defaults).
  Added keep-alive agents + 600s requestTimeout so proxy/conform master
  downloads (hundreds of MB) don't stall and leave assets in 'processing'.

services/mam-api/src/s3/client.js

@@ -2,8 +2,20 @@ import { NodeHttpHandler } from '@smithy/node-http-handler';
 import { S3Client, GetObjectCommand, DeleteObjectCommand, HeadBucketCommand, ListObjectsV2Command } from '@aws-sdk/client-s3';
 import { getSignedUrl } from '@aws-sdk/s3-request-presigner';
 import { Upload } from '@aws-sdk/lib-storage';
+import http from 'node:http';
+import https from 'node:https';
 import pool from '../db/pool.js';
 
+// Dedicated keep-alive agents with a high socket ceiling. Without these the
+// SDK uses Node's default agents (effectively short-lived, low reuse); when the
+// API proxies media (/video, /hls pipe the full S3 body through Express) those
+// long-lived streaming sockets starve control-plane calls (DeleteObject, the
+// proxy worker's master download), which then time out → assets stuck in
+// 'processing', "s3 delete failed", and dead browser playback. A large pool +
+// keep-alive lets streams and control ops coexist.
+const _s3HttpAgent  = new http.Agent({  keepAlive: true, maxSockets: 256, maxFreeSockets: 32, timeout: 120_000 });
+const _s3HttpsAgent = new https.Agent({ keepAlive: true, maxSockets: 256, maxFreeSockets: 32, timeout: 120_000 });
+
 // ── Mutable config ────────────────────────────────────────────────────────────
 let _cfg = {
   endpoint:  process.env.S3_ENDPOINT   || '',
@@ -23,9 +35,17 @@ function buildClient(cfg) {
       secretAccessKey: cfg.secretKey,
     },
     forcePathStyle: true,
-    // Hard request/connection timeouts so a stalled RustFS GET can't hang the
-    // /video and /hls endpoints forever (the original browser-playback hang).
-    requestHandler: new NodeHttpHandler({ requestTimeout: 30_000, connectionTimeout: 10_000 }),
+    // Keep-alive agents (above) prevent socket starvation between media streams
+    // and control-plane ops. requestTimeout is generous so the proxy worker's
+    // full-master download (hundreds of MB) doesn't abort mid-transfer and leave
+    // the asset stuck in 'processing'; connectionTimeout stays short so a dead
+    // endpoint fails fast rather than hanging /video.
+    requestHandler: new NodeHttpHandler({
+      httpAgent:  _s3HttpAgent,
+      httpsAgent: _s3HttpsAgent,
+      requestTimeout: 300_000,
+      connectionTimeout: 10_000,
+    }),
     requestChecksumCalculation: 'WHEN_REQUIRED',
     responseChecksumValidation: 'WHEN_REQUIRED',
   });

services/worker/src/s3/client.js

@@ -1,8 +1,11 @@
 import { S3Client, GetObjectCommand, PutObjectCommand } from '@aws-sdk/client-s3';
+import { NodeHttpHandler } from '@smithy/node-http-handler';
 import { createReadStream, createWriteStream } from 'fs';
 import { readdir } from 'fs/promises';
 import { join, extname } from 'path';
 import { pipeline } from 'stream/promises';
+import http from 'node:http';
+import https from 'node:https';
 
 const CONTENT_TYPES = {
   '.m3u8': 'application/vnd.apple.mpegurl',
@@ -10,6 +13,14 @@ const CONTENT_TYPES = {
   '.mp4':  'video/mp4',
 };
 
+// Keep-alive agents + a long request timeout. The proxy/conform jobs download
+// full master files (hundreds of MB) and upload HLS segments; the SDK defaults
+// (no keep-alive, 0/short timeouts under contention) caused master downloads to
+// stall and abort, leaving assets stuck in 'processing'. Generous timeout +
+// pooled sockets make these large transfers reliable.
+const _httpAgent  = new http.Agent({  keepAlive: true, maxSockets: 128, timeout: 600_000 });
+const _httpsAgent = new https.Agent({ keepAlive: true, maxSockets: 128, timeout: 600_000 });
+
 const createS3Client = () => {
   return new S3Client({
     region: process.env.S3_REGION || 'us-east-1',
@@ -19,6 +30,12 @@ const createS3Client = () => {
       secretAccessKey: process.env.S3_SECRET_KEY,
     },
     forcePathStyle: true,
+    requestHandler: new NodeHttpHandler({
+      httpAgent:  _httpAgent,
+      httpsAgent: _httpsAgent,
+      requestTimeout: 600_000,
+      connectionTimeout: 15_000,
+    }),
   });
 };