[BUG] Single prompt on iOS/iPadOS spawns multiple sessions that all execute independently, multiplying quota consumption

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

On the Claude iOS/iPadOS app (Code tab), submitting a single prompt silently spawns multiple duplicate sessions — typically 2, sometimes 4 — each created with the same original prompt.
Critically, these are not empty duplicates in the session list: each duplicate actually executes the prompt independently. I can read the responses, and they are different from one session to another, confirming that each is a real, separate agent run consuming tokens.
The result is that a single intended prompt burns 2–4× the expected quota, with most of that work discarded.
This is distinct from the following, which describe different failure modes:
• #31685 (web: new session creates two identical sidebar entries — but those are duplicates in the list, not independently executing runs)
• #27658 (desktop: multiple sessions cross-contaminate to run the same last prompt — my sessions run the same original prompt but produce divergent outputs)
Impact
• Quota / usage is consumed at 2–4× the expected rate for affected prompts.
• No error is shown; the multiplication is silent.
• The duplicated runs produce divergent output, so there is no single authoritative result.

What Should Happen?

A single submitted prompt should create exactly one session and trigger exactly one execution, consuming quota once.

Error Messages/Logs

Steps to Reproduce

Steps to Reproduce (trigger confirmed: app backgrounding mid-execution)

1. Open the Claude app on iPadOS (iPad Pro 12.9” 5th gen, iPadOS 26.5).
2. Go to the Code tab and submit a single prompt to a Cloud session.
3. Wait until Claude Code has clearly started executing — it acknowledges the task and begins taking actions (in my case it had been running ~1–2 minutes).
4. Switch to another app (send the Claude app to the background).
5. Return to the Claude app.
6. Observe that additional sessions have been launched — in my case, 3 extra sessions (4 total), each independently executing the same original prompt and producing divergent output.

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

1.260702.3

Platform

Anthropic API

Operating System

Other

Terminal/Shell

Other

Additional Information

Note: I remained on-screen through submission and initial execution; the multiplication occurred specifically after the background → foreground cycle, not at submit time. This points to the reconnect path re-issuing the prompt as a new executing session rather than re-attaching to the already-running one.
Actual Behavior
A single submitted prompt creates 2–4 sessions, each independently executing the prompt and each consuming quota, with divergent outputs.
Environment
• Device: iPad Pro 12.9” (5th generation) [also occurs on iPhone? — unconfirmed]
• iPadOS version: 26.5
• Claude app version: 1.260702.3 (updated same day; bug still present)
• Model: Opus 4.8
• Session type: Claude Code Cloud session (created via “+” in the Code tab)
• Confirmed trigger: backgrounding the app while a session is mid-execution, then returning to foreground.
• Frequency: Every time — reproduces 100% when creating a session from scratch and backgrounding during execution.