Hard to know because it isn't necessarily a specific penalized targeted term being 950'd. As in a targeted term 950'd but other terms still rank. Pretty much nothing ranks and seen many instances where targeted and non targeted stuff will bring a page up 950'd. It has been ever changing for us so it is hard to pinpoint while also acting somewhat differently from page to page more than likely due to other factors. I do want to mention that this is site wide. Across many thousands of pages and different styles of writing and structure. In other words this is pretty vast.
Here is a format that we use for our site and the common elements of each page:
Meta Titles: All unique in that they have different title text. Some words and keywords can be found in others. When you have a section on widgets you will have multiple titles with wigets in it. But not all. and sometimes do and don't contain targeted phrases.
Some titles do and don't contain all keywords targeted.
The title is usually the same as h1 Tags. Have the ability to vary it if we want. But makes no sense for articles to do so. Even at that this is a very common practice for those to match p. If the titles are long we will shorten them up and leave the longer version for the h1 headings
Example of our titles would be:
Do You Know the 4 Tips That Will Help Your Child Read Sooner?
Keywords: Gone so no worries there.
Descriptions: Vary and not necessarily contain exact phrases. Sometimes split sometimes not. Sometimes just a fraction of the phrase. Some contain more complete version of our target if the title/h1 do not contain all that we wanted. No description is the same and webmaster tools reports one short description.
h1 tags: Most generally the same as the title. We are experimenting with different variations between the h1 and title. One thing we could test is drop a keyword out of a phrase completely. Our other tests have used same key phrases but mixed around a bit.
Navigation: Before redesign it was keyword heavy but really not that bad. I see far worse now from better known sites to junk sites listed in the top 50. -- We are completing our changeover to a new navigation system where we won't repeat if possible. It does deepen the site a bit. The link text in the navigation is different than the titles of the page being linked to but yes they may contain the same keyword. Many times how can you not. Lets say I have a section called Recipes. I will probably use that as the navigation link text and use a more descriptive title like "Recipes For All Occasions" (just an example off the top of my head). In the h1 tag of the Recipes page...Well ya know you are going to use the term again. Again experimenting with variations.
h2/h3 tags: Use them when necessary. Some contain the keywords and some don't. Sometimes intentional sometimes not.
b/i/u: Never really use any of these to highlight keywords.
Related Discussions and Articles: Those are on the bottom of articles/discussions. Links only and no descriptions of articles. Yes some may contain keywords. If you are in a widget recipe section you will have titles from related pages that will have widget or recipe in it. You just can't help it much. To vary such titles would require a separate link text for all pages that show that related article designed specifically for that page. Again this is common practice especially database driven sites and helps keep visitors. Can't help discussion link text as those are user generated. Article links are the title of the article.
Footer links: just home/about us/privacy/copyright nofollowed.
We also, but not always, use words and keywords in URL's. Now we are looking hard at this because if you STRIP all html out of the page you can see repetition you normally wouldn't see viewing the page through a browser. Articles however my or may not contain enough keywords in urls (found in links on those pages) to really make a difference. Now comparing keywords in URLs to title and h1 and on page yes can be similar - On the other hand our discussions do NOT have keywords in the URL. I can change this for articles which would reduce keyword repetition in an instant (just change or scripts a bit) then we fall into NEW url across the entire site then go back to 301 redirects to compensate for all the links to our site. If it is a penalty or filter how much would pass through. This would be a last resort after the design slow roll changeover and re-bulking of new original content.
Links to our site: Natural over the years. Varies from crap to quality. Multi-topic site so links and anchors vary to the home page but more related links to sub-pages.
Do a search for a 3 word keyphrase. Count how many times in the top 200 you will find a site that has on pages factors of Keywords in url + title + Description + keywords + h tags + various link text + urls. You will find quite a lot. Thats just the top 200.
To make matters worse now - Caches of our pages are dropped. Google know about the pages but no caches. No caches = no listing which then = can't really test anything. At first I thought was because of thinning out our content but this is site wide not just new design. - To point out Google crawls our site everyday.
We don't cloak, don't use sneaky redirects, don't buy links, sell links, trade links, nor use link schemes, we do have a mixture of submitted content and our own, no cookie cutter affiliates, don't duplicate content on other sites we own (none), no hidden text, no virus or malware, no sub domains, no hidden links, no nothing. Headers are fine. www/non www redirect has been in place for years. Redesign years ago that had URL changes (removed the 301s last year as they were not needed any longer). Use a nofollow on submitted articles but not articles we write for fear of "gaming" for letting PR flow through links on those articles that may or may not be submitted elsewhere.
Really can't think of anything else. Oh yeah and virtually all of our site validates CSS/HTML 4.01 Strict both old and new design (old has a few warnings but nothing bad and still passes). Just can't think of anything THAT BAD that we have done do get a spanking like this with a big paddle having 950+ nails in it.