I have an argument called '--verbose' in the main parser. 
Then I add a subparser called 'command', to which I add an argument with the same name '--verbose' (See attached script).

When I process these args, I cannot decide afterwards if the user called 'myscript --verbose command' or 'myscript command --verbose'. I always get the same Namespace object returned from parse_args.

IMHO the Namespace object should also provide informations about the source parser, to make main parser args and subparser args distinguishable. Also if I call 'myscript --verbose command --verbose', I should get both args in the resulting Namespace object.

The Namespace object contains a combined list of all of the arguments, from both the main parser and the subparser.  I don't see any way to resolve identically name augments without breaking a lot of code that relies on the current behavior.  I would instead propose that anyone needing to distinguish between identically named args in nested parsers use distinct names as the primary name of the argument, and the non-distictive name as an alias.  For example:
    parser.add_argument('--verbose(main)', '--verbose', ...)

Running the attached file produces the following output.

# using non-distinctive args
['--verbose', 'command'] => Namespace(command='command', verbose=True)
['command', '--verbose'] => Namespace(command='command', verbose=True)
['--verbose', 'command', '--verbose'] => Namespace(command='command', verbose=True)

# using distinctive args
['--verbose', 'command'] => Namespace(command='command', verbose(main)=True, verbose(subcommand)=False)
['command', '--verbose'] => Namespace(command='command', verbose(main)=False, verbose(subcommand)=True)
['--verbose', 'command', '--verbose'] => Namespace(command='command', verbose(main)=True, verbose(subcommand)=True)

One change and one minor problem with my suggestion.

First, you don't need the second alias, since it's a prefix of the argument name.  Also, HelpFormatter._format_actions_usage ends with a bit of code labeled "clean up separators for mutually exclusive groups"; it removes matched parentheses from the argument names.  :(  I'd recommend that you instead use one of the following (all of which I've tested):

    parser.add_argument('--verbose[main]', ...)
    parser.add_argument('--verbose{main}', ...)
    parser.add_argument('--verbose<main>', ...)

I don't think there's any easy way for argparse to automatically do what you want. However, there's a simple workaround - just use the dest= argument and specify two different destinations:

>>> parser = argparse.ArgumentParser()
>>> parser.add_argument('--verbose', action='store_true', dest='verbose_main')
>>> subparsers = parser.add_subparsers(help='command', dest='command')
>>> cmd1_parser = subparsers.add_parser('command')
>>> cmd1_parser.add_argument('--verbose', action='store_true', dest='verbose_sub')
>>> parser.parse_args(['command', '--verbose'])
Namespace(command='command', verbose_main=False, verbose_sub=True)
>>> parser.parse_args(['--verbose', 'command'])
Namespace(command='command', verbose_main=True, verbose_sub=False)

Closing as a wontfix, but feel free to reopen if you want to provide a patch for something.

If you'd like to help improve the documentation so that the workaround for your problem is more obvious, please contribute to Issue 15428.