The traffic does not return back to restarted TiKV node

## Bug Report

When there is a range select queries on replica like s

### 1. Minimal reproduce step (Required)

1. In a cluster with zonal labels run range queries read workload only in one zone:
```
sysbench oltp_read_only --mysql-ignore-errors=1213,1020,1205,1105,9005 --point_selects=0 --range_selects=true --range_size=2 --simple_ranges=1 --sum_ranges=0 --order_ranges=0 --distinct_ranges=0 --distinct_ranges=0 --skip_trx=true --tables=1 --table_size=1000000 --threads=1 --mysql-host=${HOST} --mysql-user=root --mysql-db=test --mysql-port=4000 --report-interval=1 --time=900 run
```

It is important to run only reads with coprocessor tasks.

2. Restart TiKV in the same zone as traffic generated from.

3. Wait for TiKV node come back.

### 2. What did you expect to see? (Required)

The traffic should return to TiKV node after restart.

### 3. What did you see instead (Required)

The traffic never returns to the restarted node until first point get read is issued

### 4. What is your TiDB version? (Required)

8.5.2