Retention of frontend state upon reloading the GUI mid-generation

[c0dezer019]

Is there an existing issue for this?

• I have searched the existing issues and checked the recent builds/commits

What problem are you experiencing?

Reloading the GUI in ComfyUI_frontend during an active generation causes the user interface to lose sync with the current job and node progress. TAE (Task/Activity/Event) preview no longer updates, and the UI cannot properly restore precise ongoing task/progress state until completion or restart. Restarting a long-running job can significantly impact work, especially if waiting until job completion results in a bad outcome that could have been stopped before it finished. And due to loss of TAE preview, I'd have no way of knowing if it was going right or going wrong, making cancelling a gamble.

Context: I'm on an older computer that requires running in CPU mode, since GPU does not have support for pytorch. Therefor, gens are usually significantly longer than average, so it's not merely just a matter of going "Well, I'll just wait the 4 minutes and see what I get."

When does this problem occur?

• This occurs whenever the web UI is refreshed while a job is actively generating.
• The UI immediately shows active tasks in the queue but does not regain proper tracking of ongoing node execution or resume live TAE updates.
• Impacts monitoring or debugging of long-running jobs, especially in cases of accidental page reloads, or a reload becomes necessary mid-generation.

How often do you encounter this problem?

Multiple times per day

How much does this problem affect your workflow?

Significantly slows down my work

Current workarounds

Monitor progress through the task queue, but have to wait for ongoing jobs to finish before status tracking is correct again. No way to restore live node progress/TAE preview mid-job, which may mean loss of context or missing issues.

Ideas for solutions (Optional)

Perhaps the UI could fetch the current node/task/TAE state from the backend and provide full state restoration after a reload, minimizing operational blind spots.

Additional context

• Reproducible in multiple browsers.
• Users relying on live feedback/progress through TAE lose visibility after reload.
• No browser console errors are shown for this condition.
• Example: Start a generation, reload page, progress/TAE lost until job ends.
• May impact collaborative or batch automation workflows if UI state is lost during processing.

┆Issue is synchronized with this Notion page by Unito

[coderabbitai]

🔗 Related PRs

#7394 - fix: work around Chrome GPU bug causing severe lag when dragging links [merged]

---

Enable issue planning

_{To enable issue planning, add the following to your .coderabbit.yaml:}

issue_enrichment:
  planning:
    enabled: true

_{You can then request a plan by commenting @coderabbitai plan on any issue.}

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!

[dosubot]

This is a known and unresolved limitation in ComfyUI_frontend. When you reload the GUI during an active generation, the frontend loses sync with ongoing job and node progress because execution state (like node progress and TAE preview) is only tracked in memory and not restored from the backend after reload. The UI will reconnect to the backend and show the current queue, but it does not regain live progress or TAE updates for jobs already in progress—only new prompts trigger state updates and visualization. This issue is tracked in several places, including #6430, #7961, and #3856.

The backend does expose APIs to fetch the current queue, job details, and history, but the frontend does not currently use these to restore live node/task/TAE state after a reload. Fixing this would require the frontend to query the backend for in-progress jobs after reconnecting and repopulate its execution state so the UI can resume showing progress and previews. Right now, there is no workaround for restoring live progress mid-job; only the queue and history are refreshed after reload, and live updates are lost until the job completes or a new prompt is submitted.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}