Hi Toggl team,
Great product as always! Thank you.
Since approximately Monday July 6, the Track web app (focus.toggl.com) became nearly unusable for me on some machines: multi-second click response, “this page is slowing down your browser” warnings, tab pinning a CPU core, and consoles full of interrupted connections to wss://focus.toggl.com/websockets and failed requests to api.plan.toggl.com (status null).
I traced the root cause: disabling HTTP/3 in the browser (Firefox: network.http.http3.enable = false) instantly restores normal performance. Plain TCP connections to your hosts (curl) always worked, with valid certificates. So the failure mode is: browsers negotiate HTTP/3 over QUIC (UDP 443) to your Google Cloud frontends, something on the network path mangles that UDP traffic mid-connection rather than rejecting it cleanly, and browsers hang/retry instead of falling back to TCP.
My question: did you enable or expand HTTP/3 / QUIC advertisement (Alt-Svc) on your load balancers around that date? A server-side rollout would exactly explain a sudden onset for users behind networks or security stacks that mishandle QUIC — a fairly common enterprise configuration.
Maybe this is a one-off from my setup or location; or else it might explain any other “app suddenly slow” complaints you are receiving this week. The client-side workaround (disable HTTP/3) is effective, and affected users’ network admins can fix it properly by making their QUIC blocking reject handshakes cleanly, which restores instant TCP fallback.
While investigating, I also noticed that crisp chat service was inducing a 2-3 second page-load latency, so I have blocked it and regained additional performance! Perhaps a setting to turn chat on and off would be useful; just a thought.
Happy to provide HAR captures, console logs, or timestamps if useful.
Browsers affected: Firefox-based (Zen, Firefox, Edge), Linux and Windows