I'm a little confused about subscribe_on and the 'scheduler' parameter of Observable.create:

Example

import rx
import rx.scheduler
import rx.operators as op
import time
import threading

scheduler = rx.scheduler.ThreadPoolScheduler(max_workers=5)


def producer(observer, scheduler):
    print("Provided scheduler {} - Current Thread: {}".format(scheduler,
                                                              threading.current_thread().name))
    observer.on_next(scheduler)
    observer.on_completed()


obs: rx.Observable = rx.create(producer)


def handle(s):
    if (s == scheduler):
        print("Worked")
    else:
        print("Failed")


print("First Option")
obs.pipe(
    op.subscribe_on(scheduler=scheduler)
).subscribe(lambda x: handle(x), scheduler=scheduler)
time.sleep(0.25)


print("Second Option")
obs.pipe(
    # op.subscribe_on(scheduler=scheduler)
).subscribe(lambda x: handle(x), scheduler=scheduler)

time.sleep(0.25)
print("Third Option")
obs.pipe(
    op.subscribe_on(scheduler=scheduler)
).subscribe(lambda x: handle(x))

time.sleep(1.0)

Options from example

I would expect...

1. that the passed in scheduler is the ThreadPoolScheduler (but is None) and the subscribe logic is executed by a ThreadPool-Worker (True)
2. that the passed in scheduler is the ThreadPoolScheduler (True) and the the subscribe log is executed on any thread (in this case its the MainThread but maybe it would also be more logical to run this on the subscribe call scheduler (ThreadPool)?)
3. Somehow i would also expect that the passed in scheduler is the ThreadPoolScheduler (which it is not) and I can see why, but then I'm asking myself how the "scheduler" parameter can be used? The subscribe logic runs correctly on a Worker thread

I guess it all resolves around understanding what the scheduler parameter that the subscriber of Observable.create gets really represents.

I hope you can understand my problem

rx==3.1.1
python=3.8.3-64bit Windows

Cheers
Jascha