Strange performance issues on windows

[sh1nkey]

Did some benchmarks

Code example:

def timit(func):
    async def wrapper():
        begin = time.time()
        await func()
        stop = time.time()
        print(f"total time of {func.__name__}: {stop - begin}")

    return wrapper

@timit
async def natspy_main_nonconcurrent() :
    client = NATS()

    await client.connect('downtimer-nats1:4222')
    
    for _ in range(1_000_0):
        await client.publish("some", b"1")

    await client.close()


@timit
async def natsrpy_main_nonconcurrent() :
    client = Nats(
        addrs=['downtimer-nats1:4222']
    )

    await client.startup()

    for _ in range(1_000_0):
        await client.publish("some", b"1")

    await client.shutdown()

Output:

total time of natspy_main_nonconcurrent: 0.02205514907836914
total time of natsrpy_main_nonconcurrent: 2.494574546813965

Additional details:

• ran on Windows 10
• cProfile of natspy run shows 263970 function calls, cProfile of natsRpy shows 990367 function calls
• when ran with bigger payload (2 million symbols), natsRpy becomes faster then natspy not by a big margin
• running benchmark in Dockercontainer doesn't really change anything
• with using gather, function call count becomes somewhat similar, though perfomance is also strange (natspy 0.05 vs natsRpy 0.45)