Two graphs made using the information we collected from the first twelve sessions of the Lost City game. 
(I like to geek out with this sort of analysis, and it adds to my enjoyment of the game. I know folks who see it as trying to dissect a butterfly to find the beautiful part. Skip it if you wish.)
A couple thoughts on the first graph.
The one-session column will always be large since there are people who play one session and do not return. Other players like to play a new character each session.
The mortality bulge in the second/third sessions is largely due to probability. The majority of characters in any one of these sessions have attended only 2-3 previous sessions; by the twelfth session the average character had attended 2.7 sessions. (This is changing, the average “session age” of the group is rising, fast, and I expect mortality to drift upwards on the chart along with it.)
There were also learning curve issues: we lost a bunch of characters in the first few sessions as we learned. Hard to tell if it was learning the particulars of this dungeon, or learning about the game in general. 
Sternum also raised an excellent point:
When I roll up a fresh, level one character, I tend to play a lot more recklessly than when I’m playing a seasoned adventured because I have little to lose. I went through three guys in Tavis’ game one evening because, for those few hours after a character is created, death wasn’t really much of a threat because an equivalent (or better) character would only cost me a minute or two of dice rolling.
This is especially powerful in white box, or any early D&D where random things like opening the wrong door can kill a low-level character. Newly crafted characters are more likely to volunteer… 
I think we will have more deaths in the next few sessions as we move into more lethal areas (teleport traps, efritt, astral monsters, and the beast lord) without having gained new informational resources. Wear your (low, soft) running shoes.
The second graph is easier to read: I see no linear relationship between sessions played and experienced gained. The majority of experience comes from treasure (as it should!) and the amount of loot we get has been wildly variable. Your best bet is to show up for as many games as possible and hope you attend one where we find a nice hoard. White box advancement is just not a direct result of “putting your time in.” It has more to do with adroitness, cooperation, and luck.
 A sandbox game based on the three original D&D books, with a few changes from the DM (e.g., carousing rules, use of full spell set from AD&D, a different way of approaching HPs). Original data here:
 The crunchy part of me is obliged to point out that the data set here is too small to provide anything in way of statistical significance, but we will not let that stop us, will we?
 [Tavis, stop reading please] It would be interesting to have a total wipeout, resulting in a completely new party but with players who already know the game and the DM’s style; would we have as many deaths in the first few sessions of the new party? [/resume Tavis]
 To reinforce this, think about how excited everone was about Lydio’s special spider-sense ability during the last session.