fix: activation command may get retried in a different connection, or failed due to queue full or dangled due to auto flush is false

[okg-cxf]

Bug Report

1. activation command should never be retired in a different connection;
2. activation command should never fail due to queue full;
3. activation command should never get dangled in writeBuffer even if auto flush is false.

Refer to the MR for more details: https://github.com/redis/lettuce/pull/2962/files

Current Behavior

Stack trace

// your stack trace here;

Input Code

Input Code

// your code here;

Expected behavior/code

Environment

• Lettuce version(s): [e.g. 5.0.0.RELEASE, 4.2.2.Final]
• Redis version: [e.g. 4.0.9]

Possible Solution

Additional context

[tishun]

Hey @okg-cxf ,

could we take a step back and talk a bit more about the use case you are handling?

Can you give a short example of the issue you are trying to solve? When does it occur?
Is there an easy way to reproduce it?