[Feature Request] Implement performance monitoring and SLA guarantees for Claude Max subscription tier

Bug Description
Between April and June 18, while using the Claude Max x20 plan, I experienced significant inconsistency and frustration. During multiple periods, there was a noticeable degradation in performance: lower-quality responses, difficulty following instructions, excessive token consumption, and frequent need to repeat or re-explain requests. In many cases, Claude failed to deliver what was explicitly requested, which negatively impacted my productivity and workflow.

More recently, I have seen some improvements, and my usage was restored on one occasion, which I appreciate. However, these recent improvements do not fully offset the poor experience I had over the past several months.

As a power user who relies on Claude for complex technical, engineering, automation, and software development tasks, I expect consistent reasoning quality, reliable instruction-following, efficient token usage, and stable performance. My main feedback is that reliability and consistency need significant improvement. A premium plan should provide a dependable experience over time, not fluctuate between excellent and poor performance.

I genuinely want Claude to succeed because, at its best, it is an exceptional tool. However, the issues I experienced between April and June reduced my confidence in the platform and made it difficult to justify the premium subscription during those periods.

Environment Info

• Platform: linux
• Terminal: windows-terminal
• Version: 2.1.181
• Feedback ID: b90546d1-8460-41f2-8288-cf458c69edbc

Errors

[]

Summary

Claude Code / Loop does not continue unattended recurring runs correctly.
When I configure a loop to run multiple iterations (for example, every 1 hour for 24 hours), it repeatedly asks whether I want to run in cloud or local mode on each sequence. Instead of continuing the scheduled loop, it stops and waits for a new user confirmation every time.

This breaks unattended automation and makes recurring runs unusable.

Expected behavior

If I define a recurring loop with multiple iterations, Claude Code should:

• keep running through the configured iterations automatically
• preserve the selected execution mode for the whole loop
• only ask for a mode selection once, at the start, if required
• not interrupt every iteration with the same question

Actual behavior

Claude Code repeatedly prompts me to choose between cloud and local on every sequence.
It does not continue the loop autonomously and waits for my input each time.

Example

I asked for a loop like:

• run every 1 hour
• continue for 24 hours
• total of 24 executions

Instead of completing the loop, it keeps stopping and asking again whether to use cloud or local execution.

Impact

This breaks:

• unattended scheduled runs
• long-running automation
• recurring workflows
• batch execution
• reliability of Loop / scheduler behavior

Additional context

This feels related to the recurring-task / loop execution state not being preserved across iterations.

Environment

• Platform: Linux
• Terminal: Windows Terminal
• Version: 2.1.181

Repro steps

1. Start a loop or recurring task with multiple iterations.
2. Configure it to repeat over time, for example every 1 hour for 24 hours.
3. Observe that Claude asks again whether to run in cloud or local mode on each cycle.
4. The loop stops and waits for my confirmation instead of continuing automatically.

Expected result

A recurring loop should continue through all iterations without repeated mode prompts.