[lena_punkt]

Okay, I found it and want to share my solution with others who might have the same problems:

You can set the variable "simultaneous_downloads" within a recipe to 1. Maybe you need to add it. example:

from calibre.web.feeds.news import BasicNewsRecipe

class BasicUserRecipe1257093186(BasicNewsRecipe):
    title          = u'Blogs'
    oldest_article = 14
    language = 'de'
    no_stylesheets = True
    max_articles_per_feed = 100
    remove_tags = [dict(id=['respond', 'ackt', 'nav', 'commentblock',
        'l_sidebar', 'r_sidebar']), dict(attrs={'class':['widgettitle',
            'bt-links', 'navigation']})]

    feeds          = [(u'Hagen - DerWesten', u'http://www.derwesten.de/staedte/hagen/rss'), (u'Tagesschau', u'http://www.tagesschau.de/xml/rss2')]
    multithreaded_fetch = False
    simultaneous_downloads = 1

It did not help that much with the RAM problem, but seems a bit faster on my machine.