When jemallocator::Jemalloc is used as #[global_allocator] and ran with cargo run, the stacksize 128 of "flexi_logger-flusher" thread seems not enough, reporting thread 'flexi_logger-flusher' has overflowed its stack error. When ran with cargo run --release, it does not have this problem.
Here is the code of main.rs :

use std::thread;
use std::time::Duration;

use flexi_logger::{LoggerHandle, WriteMode};

#[global_allocator]
static ALLOC: jemallocator::Jemalloc = jemallocator::Jemalloc;

fn main() {
    let _handle = init_logger();
    thread::sleep(Duration::from_secs(1000));
}

pub fn init_logger() -> LoggerHandle {
    flexi_logger::Logger::try_with_str("info")
        .unwrap()
        .log_to_stdout()
        .write_mode(WriteMode::BufferAndFlush)
        .start()
        .unwrap()
}

And this Cargo.toml:

[package]
name = "practice-flexi"
version = "0.1.0"
edition = "2021"

[dependencies]
log = "*"
flexi_logger = { version = "0.19", features = ["async"] }
jemallocator = "*"

This test program is built with rustc 1.57.0-nightly (5b210643e 2021-10-11)

Improving the stacksize to at least 60*1024 on my machine, the problem can be sovled.