Remove viem websocket closing hackery when viem is updated

[rvagg]

wevm/viem#4378 is closed by wevm/viem#4601; when we inherit that we should be able to remove the extra process.exit stuff we're doing for unclosable websockets because this was precisely the issue we encountered here.

[SgtPooki]

2026-06-16

Dug into whether the upstream fix has landed and whether a bump alone closes this.

The viem fix is in: wevm/viem#4601 shipped in viem 2.50.0. Changelog: "Fixed socketClient.close() triggering an unwanted reconnect loop that prevented the Node.js process from exiting." That's the reconnect-on-close half of the problem at src/cli.ts:79-84.

The bump is clean and overdue. @filoz/synapse-sdk@1.0.1 and @filoz/synapse-core@0.7.0 already declare viem: ^2.52.0, but our lockfile pins viem 2.48.4, which doesn't even satisfy that range. A refreshed install pulls 2.52.x (our own ^2.48.3 allows it too).

But a bump alone won't let us delete the force-exit. The second cause still stands: for bare-address accounts (read-only and session-key CLI modes), src/core/synapse/index.ts:248 wraps the transport in custom({ request: resolved.request }), which keeps only .request and drops the websocket socket. So we have no handle to call close() on, regardless of the viem fix.

Remaining work to close this:

1. Bump viem (folds into the synapse ^2.52.0 requirement).
2. Retain the websocket socketClient so we can reach it at shutdown.
3. Call close() on CLI exit (now safe post-2.50.0) and remove the force-exit at src/cli.ts:85-94.
4. Verify CLI commands drain the event loop and exit cleanly without the force-exit.

[BigLep]

Triage comment from 2026 FOC Board Triage (Post GA): easy. can do now AFAIK