fix(cli): stop telemetry hangs and errors on blocked networks

[pamelachia]

On networks that block or blackhole eu.posthog.com (corporate proxies, CI and Docker egress policies, air-gapped machines), telemetry breaks the CLI itself: stable v2.109.1 pays a 5s trailing hang and exits 1 with an UnknownError payload after every command, long-running commands print raw SDK stack traces to stderr, and the Go binary's shutdown wait is unbounded. The only workaround was supabase telemetry disable, which permanently removes those users from CLI analytics. This PR makes telemetry strictly fire-and-forget: a failed delivery never changes a command's output, exit code, or runtime beyond a 2s cap.

Changed:

• Exit 1 after successful commands: the TS shutdown finalizer was Effect.promise(...).pipe(Effect.ignore), but Effect.promise turns the shutdown-deadline rejection into a defect that Effect.ignore cannot swallow, so a telemetry timeout failed the whole command. The rejection is now caught on the promise itself.
• Unbounded and stacking exit delays: one shared 2s budget (EXIT_DELAY_CAP_MS) bounds each request and the entire _shutdown drain on the TS side, and the Go client gets the same 2s ShutdownTimeout (the SDK default waits indefinitely, including retries). The previous 5s shutdown deadline left room for two sequential 2s request timeouts (~4s exits).
• Work outliving the deadline (review round 3): the SDK's shutdown deadline only stops the wait, it does not cancel work. The in-flight request kept running past scope release and the drain then started the queued request, keeping the runtime alive ~4s instead of 2s. A lifecycle AbortController now composes into every request's signal and fires when the deadline hits, so no request survives or starts after the scope closes. Covered by a unit regression asserting zero active requests one tick after release (red on the previous code) and an e2e test that runs the compiled binary against a TCP blackhole and asserts wall-clock process exit, empty stderr, and that telemetry actually reached the wire.
• Stderr noise: Go routes SDK logs to the --debug logger; TS reports delivery failures as delivered inside the wrapped fetch, because posthog-node has no logger hook, only hardcoded console.error calls, so an infallible transport is the one supported way to get no retries, no output, and no stall.

Note: the exit bounds are caps, not floors. On healthy networks both shells return as soon as the flush completes (flushAt: 1 keeps the send overlapping command runtime), and fully detaching the send instead would race process exit and silently drop events on healthy networks. The accepted tradeoff is that deliveries slower than 2s are dropped; telemetry is best-effort, observable under --debug on the Go side only.

Linear

• fixes GROWTH-1001

[github-actions]

Supabase CLI preview

npx --yes https://pkg.pr.new/supabase/cli/supabase@74a0e053158089fae8129f41d30300d2a1e0dbce

Preview package for commit 74a0e05.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: a96c6f5826

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: e38cb95741

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[avallete]

@codex review

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Already looking forward to the next diff.

Reviewed commit: 77ce8398ee

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[avallete]

The Go-side timeout and debug-log routing in this PR look good and should stay. I don’t think we need to additionally coordinate one shared shutdown budget across TS parents and Go children: that would broaden the change into cross-runtime telemetry ownership for a transitional sidecar. Let’s keep the blocking scope focused on ensuring the TS client’s 2s shutdown cap also bounds actual Bun process exit.

[pamelachia]

also if this doesn't work, and you have a preferred method @avallete that isn't how i did it, can i close this PR and I'll create a linear issue for CLI team to figure this out?

[avallete]

All good, thank's for all the adjustments you made through the review loops @pamelachia !