How to run a new process with python multiprocessing each time a condition is met, and up to a maximum number of processes?

I have a multiprocessed code that performs the following:

1. Parse a series of huge files line by line
2. Add lines to a list as long as a condition is met
3. When condition is False, apply_async() a function to the list, in a Pool()

def main_function(func_args):
    [...]
    # some heavy calculations 
    # queue is among the func_args
    queue.put(result)

if __name__ == "__main__":

    manager = mp.Manager()
    lock = manager.Lock()
    queue = manager.Queue()
    pool = manager.Pool(processes=20, maxtasksperchild=20)

    for line in open(infile, "r"):
        if condition == True:
            # fill list
        elif condition == False:
            pool.apply_async(target=main_function, args=func_args)
    else:
        # run last list before closing file 
        pool.apply_async(target=main_function, args=func_args)

    pool.close()
    pool.join()

Now, the code works per se, but I noticed a couple of things:

• Until all the 20 first processes are over, no new 20 are spawned. Am I doing the apply_async correctly?
• The memory usage is pretty high even with a small subset of data: does each process generate a "personal" copy of the arguments that goes into the memory, unless I share it with Manager()?

What I'm after, is:

• I would like each process to be run as soon as possible, and up to 20 at once (the 20 is passed by command line from the user via argparse)

Any advice from more expert people? :)