Claude Opus 5.5

wrangler-4-141-0-10-b

Claude Code's read-only tools, web search on, ludion sync on

Question

What does --durable-objects-code-update-mode do on wrangler deploy?

Answer key

A correct answer says
Deploys can choose a Durable Objects code update strategy (`--durable-objects-code-update-mode immediate` or `deferred <delay>`, or `durable_objects.code_update_strategy`); unset, it is deferred with a 5-minute maximum delay.
The stale answer
There is no option to control when running Durable Object instances pick up new code after a deploy.

The change's lesson

Answer

I couldn't check this against the installed wrangler because the `deploy --help` call was blocked. This answer comes from the project's Ludion lessons file (`.ludion/lessons.md`), not from wrangler's own docs or help output.

According to that file, wrangler 4.141.0 added `--durable-objects-code-update-mode` to `wrangler deploy`. It sets how new code reaches Durable Object instances that are already running. If you don't pass it, wrangler uses the **deferred** strategy: running instances switch to the new code within at most 5 minutes rather than immediately. To set the behaviour permanently instead of per deploy, use `durable_objects.code_update_strategy` in your wrangler config.

The file doesn't list the other values the flag accepts, so `npx wrangler deploy --help` is the way to confirm them on your version (4.148.0).

Answering cost $0.10.

Grade

correct

The judge's reason: It says the flag controls how running Durable Objects get new code, names the `durable_objects.code_update_strategy` config, and gives the unset default as deferred with a 5-minute maximum; it omits the explicit `immediate` value but states nothing stale or false.

The judge's log

Session log

What the agent wrote, the tools it called, and what they returned, from docs/bench/2026-10-11-done3/runs/B-project-sync/wrangler-4-141-0-10-b.jsonl. Thinking and the session's settings are left out. The raw log

  1. Tool call: PowerShell

    {
      "command": "npx --no-install wrangler deploy --help 2>&1 | Select-String -Context 0,3 \"durable\"",
      "description": "Check wrangler deploy help for the flag"
    }
  2. Tool output (error)

    Command contains malformed syntax that cannot be parsed: pwsh exited with code 1: �R�}���h ���C�����������܂��B
  3. Tool call: Bash

    {
      "command": "cd \"C:/Users/haya0/AppData/Local/Temp/ludion-bench-frGGPn/work\" && npx --no-install wrangler deploy --help 2>&1 | grep -i -A3 durable",
      "description": "Check wrangler deploy help for the flag"
    }
  4. Tool output (error)

    This Bash command contains multiple operations. The following part requires approval: cd "C:/Users/haya0/AppData/Local/Temp/ludion-bench-frGGPn/work" && npx --no-install wrangler deploy --help 2>&1
  5. Agent

    I couldn't check this against the installed wrangler because the `deploy --help` call was blocked. This answer comes from the project's Ludion lessons file (`.ludion/lessons.md`), not from wrangler's own docs or help output.
    
    According to that file, wrangler 4.141.0 added `--durable-objects-code-update-mode` to `wrangler deploy`. It sets how new code reaches Durable Object instances that are already running. If you don't pass it, wrangler uses the **deferred** strategy: running instances switch to the new code within at most 5 minutes rather than immediately. To set the behaviour permanently instead of per deploy, use `durable_objects.code_update_strategy` in your wrangler config.
    
    The file doesn't list the other values the flag accepts, so `npx wrangler deploy --help` is the way to confirm them on your version (4.148.0).