I work on the technical side of two competing newspapers that used to operate under a federal Joint Operating Agreement (JOA - a very old federal law) and decided to split a couple years ago and dissolve that agreement - they randomly still share a 50-50 revenue split. My insight into your problems is rather unique because we've faced the same issues.
First off, none of the big CMS platforms work well for a news distribution model. They all suck. They aren't designed for automated content workflows and they don't scale well. The stuff you don't hear about are how IT personnel struggle under "big breaking news" items to keep servers alive and functioning to serve up the content to visitors. Despite whatever the plugin authors for major CMSes have claimed over the years, the various caching plugins for the major CMS products simply don't cut it when the entire Internet drops in for a visit. IMO, you should avoid these CMS products like the plague that they are. They sound good until you try to actually use them in a newspaper context and then they fail hard, especially under load.
Besides software, you also need to figure out hardware and bandwidth. A VPS allows you to scale up more easily (VPS.net is actually pretty good), but they get pricey fast and bandwidth is expensive per GB of transfer. We buy bandwidth in bulk, so hosting on-site is more advantageous and cheaper. Our problem as well is hardware. The newspaper industry isn't doing so hot these days, so the money for more physical server hardware just isn't there. We have to make do with what we've already got.
You might want to look into a little product called TownNews (aka Blox CMS). They've convinced a whole slew of newspapers across the nation to use their product. They can natively feed in AP content and pull content in from Quark, InDesign, and other sources into their database. Their server farm scales fairly well. It looks and sounds really slick until you start using it. The most irritating thing for our newsroom folks? Caching. Their 5 to 30 minute cache is the most infuriating thing to run into, especially with breaking news where updates are flowing in every couple of minutes. You have to wait for that cache to clear for the updates to show up. They also have usage limits of dynamic scripting resources because their IT folks have a baton stuck up their collective rears (don't try to visualize that). If your newsroom is large enough and you are behind a NAT firewall, your staff will all get HTTP 429 errors every other month because TownNews does IP rate limiting, which actually affects Google search result placement because Googlebot occasionally can't index those websites that run on their platform. However, they can weather a storm of Internet visitors and have reverse publishing options (stories can flow from their CMS into InDesign), so they are doing some things right. As an example of their robustness, a huge national story in this area not too long ago only caused parts of their server farm to go down twice for only a few minutes each time while most other papers in the area were pretty much offline the entire time (roughly 48 hours). That was the only time I've ever been impressed with the platform's caching mechanisms. Under normal load (the other 99.99% of the time), the whole caching scheme is a pain to work with from a behind-the-scenes perspective. Of course, caching plugins for the major CMS products are in a similar boat.
Basically, whatever product you end up using, something will be infuriating about it. There really isn't a good solution that's based on free software and actually scales under load. Newspaper production is a highly specialized area of expertise despite it not sounding like it from a lay-person's perspective. "Highly specialized" is something that doesn't work well with "open source". Quark and InDesign are fine until you start trying to export stuff from it onto the web or do the reverse pub thing - something they weren't really built to do. There really isn't a good open source solution, which is why specialized places like TownNews crop up and skewer you for money because they know they can.
IMO, despite their shortcomings, TownNews is probably the way to go. However, if you still want to go the open source route, look for a good caching plugin, try to find a good VPS so you can scale up fast under heavy load and has plenty of bandwidth, use Amazon AWS for serving up the static content, look for a good feed plugin (e.g. FeedWordPress is pretty good for WordPress) for importing content into the platform, hire someone to write InDesign export/import scripts, and then pray that you don't have major national stories. But hosting it yourself isn't worth the headache. I've personally run into so many issues with WordPress that it isn't funny. It simply doesn't scale under load despite what the authors continue to falsely believe - they run WordPress.com but their solution is to throw more hardware at the problem rather than face and fix the fact that their software is so poorly written that they need to start over.
One final thing: Whatever platform you choose is what you are likely going to be stuck with for a very long time. Take your time and figure out all you can about each platform so that you know all the pros and cons of each solution being sure to calculate costs based on how many people will be involved in annual maintenance time (i.e. IT salaries).