Error 500 in some data when writing data over HTTP, writable after restart

Nobody has claimed this yet.

__Steps to reproduce:__
Query data while writing, guess it's a concurrency related issue, can't happen every time.

1. 50,000 points written per second, guessing is related to cardinality(series).
2. Constantly query the written data.
3. wait, until some data can't be written.

samples:
ms_aaa_request_duration_detail,buz_env=formal,code=200,gameid=aaa,hostname=bbbb,ip=192.168.1.1,project=中文,source=/home/fee/aaa/bbbb/log/general.log,subcode=1,url=authxxx avg=46.0,max=46.0,count=1.0 1665205177446141543

__Expected behavior:__
No error and successfully write data/query data.

__Actual behavior:__
Some data queries timeout or writes timeout

__Environment info:__

* System info: Linux 4.19.0-19-amd64 x86_64
* InfluxDB version: InfluxDB v2.4.0 (git: de247bab08) build_date: 2022-08-18T19:41:15Z

__Config:__
Copy any non-default config values here or attach the full config as a gist or file.

__Logs:__
no.

__Performance:__
Generate profiles with the following commands for bugs related to performance, locking, out of memory (OOM), etc.

```sh
# Commands should be run when the bug is actively happening.
# Note: This command will run for ~30 seconds.
curl -o profiles.tar.gz "http://localhost:8086/debug/pprof/all?cpu=30s"
iostat -xd 1 30 > iostat.txt
# Attach the `profiles.tar.gz` and `iostat.txt` output files.
```
[goroutine-49-1010.txt](https://github.com/influxdata/influxdb/files/9808664/goroutine-49-1010.txt)

Start by reproducing the concurrent HTTP write/query workload described in the issue against InfluxDB v2.4.0, using the provided 50,000-points-per-second scenario. Inspect goroutine-49-1010.txt and collect the listed pprof and iostat outputs while the failure occurs. Done means writes and queries no longer time out under the reported workload.

Written by the indexing model from the issue text.