fix(client): back off event stream retries to a server that keeps failing

packages/app/src/runtime/server/client.tsx

@@ -102,7 +102,8 @@ function createServerSdkContextBase(server: ServerConnection.Any, scope: ServerS
       info(message, data) {
         if (message !== "event stream disconnected") return
         const managed = server.type === "plugin" ? server.managed : !!reconnect
-        console.info("[global-sdk] event stream disconnected", { url: transport.url, managed, ...data })
+        // Debug exports keep the console as text, where an object reads "[object Object]".
+        console.info("[global-sdk] event stream disconnected", JSON.stringify({ url: transport.url, managed, ...data }))
       },
     },
   })

packages/app/src/runtime/server/request-queue.ts

@@ -51,7 +51,8 @@ export function createRequestQueue(input: {
   // Call the browser fetch unbound; `input.fetch(...)` would make `this` the options object.
   const base = input.fetch
   const now = input.now ?? Date.now
-  const log = input.log ?? ((message, data) => console.warn(`[server-request-queue] ${message}`, data))
+  // Debug exports keep the console as text, where an object reads "[object Object]".
+  const log = input.log ?? ((message, data) => console.warn(`[server-request-queue] ${message}`, JSON.stringify(data)))
   const inflight = new Set<Entry>()
   const waiting: Array<{ entry: Entry; start: () => void }> = []
   let warned = -Infinity

packages/client/src/solid/connection.ts

@@ -31,6 +31,11 @@ export type ClientConnectionOptions = {
 
 const connectTimeout = 2_000
 const reconnectDelay = 1_000
+// Each attempt that fails before the stream connects (server offline, unreachable or rejecting us) doubles the wait
+// before the next one, up to this delay. Chromium keeps an aborted attempt's CORS preflight queued on the origin's six
+// sockets until TCP gives up minutes later, so retrying an unreachable server every few seconds piles them up. A
+// resync retries at once.
+const maxReconnectDelay = 30_000
 const connectionHistoryLimit = 50
 export const defaultIdleTimeout = 45_000
 // Longer than one server keepalive interval: a stream that is silent this long when the page
@@ -51,6 +56,7 @@ export function createClientConnection(initialApi: OpenCodeClient, options: Clie
   let flushTimer: ReturnType<typeof setTimeout> | undefined
   let stream: AbortController | undefined
   let current: AbortController | undefined
+  let pause: AbortController | undefined
   let run: Promise<void> | undefined
   let started = false
   let generation = 0
@@ -178,7 +184,12 @@ export function createClientConnection(initialApi: OpenCodeClient, options: Clie
         forced = false
         continue
       }
-      await wait(reconnectDelay, controller.signal)
+      pause = new AbortController()
+      await wait(
+        Math.min(reconnectDelay * 2 ** (attempt - 1), maxReconnectDelay),
+        AbortSignal.any([controller.signal, pause.signal]),
+      )
+      pause = undefined
     }
   }
 
@@ -208,9 +219,11 @@ export function createClientConnection(initialApi: OpenCodeClient, options: Clie
     setConnection({ status: "connecting", attempt: 0, error: undefined })
   }
 
-  // Drop the live request so the reconnect loop replaces it now instead of waiting for the idle watchdog.
+  // Drop the live request so the reconnect loop replaces it now instead of waiting for the idle watchdog. Between failed
+  // attempts, end the pause instead so the next attempt starts now.
   function resync(reason: string) {
-    if (!started || connection.status !== "connected") return
+    if (!started) return
+    if (connection.status !== "connected") return pause?.abort()
     options.log?.info?.("event stream resync", { reason, idle: Date.now() - lastActivity })
     forced = true
     current?.abort(new Error(reason))

packages/client/test/solid-connection.test.ts

@@ -1,4 +1,4 @@
-import { expect, test } from "bun:test"
+import { afterEach, expect, jest, test } from "bun:test"
 import { createRoot } from "solid-js"
 import { createClientConnection } from "../src/solid"
 import { OpenCode, type OpenCodeEvent } from "../src/promise"
@@ -59,6 +59,8 @@ function setup(input: ReturnType<typeof server>, idleTimeout: number) {
   }))
 }
 
+afterEach(() => jest.useRealTimers())
+
 async function until(check: () => boolean, timeout = 2_000) {
   const deadline = Date.now() + timeout
   while (!check()) {
@@ -163,6 +165,50 @@ test.each([
   }
 })
 
+test("a server that keeps failing before the stream connects is retried less often until a resync", async () => {
+  jest.useFakeTimers()
+  const requests: number[] = []
+  // Each resolves when the loop logs that failure, just before it starts the pause before the next attempt.
+  const disconnects = Array.from({ length: 9 }, () => Promise.withResolvers<void>())
+  const disconnected = (count: number) => disconnects[count - 1].promise
+  const api = OpenCode.make({
+    baseUrl: "http://opencode.local",
+    fetch: async () => {
+      requests.push(Date.now())
+      throw new TypeError("Failed to fetch")
+    },
+  })
+  const ctx = createRoot((dispose) => ({
+    dispose,
+    connection: createClientConnection(api, {
+      flushInterval: 0,
+      onEvent: () => undefined,
+      log: {
+        info(message, data) {
+          if (message === "event stream disconnected") disconnects[Number(data?.attempt) - 1]?.resolve()
+        },
+      },
+    }),
+  }))
+  try {
+    await disconnected(1)
+    for (const count of [2, 3, 4, 5, 6, 7, 8]) {
+      jest.advanceTimersToNextTimer()
+      await disconnected(count)
+    }
+    expect(requests.slice(1).map((at, index) => at - requests[index])).toEqual([
+      1_000, 2_000, 4_000, 8_000, 16_000, 30_000, 30_000,
+    ])
+    expect(ctx.connection.status()).toBe("reconnecting")
+
+    ctx.connection.internal.resync("Network connection restored")
+    await disconnected(9)
+    expect(requests.at(-1)).toBe(requests.at(-2))
+  } finally {
+    ctx.dispose()
+  }
+})
+
 test("a reconnect does not ask the volatile event stream to replay", async () => {
   const fake = server()
   const ctx = setup(fake, 10_000)