[alexluong]

hookdeck listen drops a healthy connection whenever the server's reply to a ping (pong) is more than about a second late. Users on slower or busier networks then see "Connection lost, reconnecting..." every few minutes.

Why: the client pings every 9s and drops the connection if no pong arrives within 10s of the previous one. That leaves about 1s of slack:

Before: ping every 9s, drop 10s after the last pong

  0s     pong received ──────────────── deadline 10s
  9s     ping sent
  10s    ✗ deadline hit, connection dropped   ("Connection lost, reconnecting...")
  10.5s  pong arrives, too late

After: ping every 10s, drop 30s after the last frame of any kind

  0s     pong received ─────────────────────────────────────── deadline 30s
  10s    ping sent
  10.5s  pong arrives ✓ deadline moves to 40.5s

	Before	After
Ping every	9s	10s
Drop after	10s without a pong	30s without any frame
Slack for a late pong	~1s	~20s
Incoming events keep it alive	no	yes
Old socket on reconnect	left open until the server cleans it up	closed

Changes (pkg/websocket/client.go):

1. Pong wait 10s → 30s; ping every 10s.
2. Any incoming frame renews the read deadline.
3. Close the old socket before reconnecting.

Tests: a busy connection with late pongs stays up; the socket closes after a read timeout; new defaults. All three fail on the old code.

🤖 Generated with Claude Code