Investigation:
Wow, that sequence is a real cement mixer! It seems to rise and
fall like the tide, with a period of about 6411/1256.
I notice a832 = 0.999999529702558, and a6411 =
0.999999996898893.
The "high tides" also follow a cycle with a non-integral
period of about 6411/131.
I would make a conjecture that for any e, an
n can be found such that 1-an < e.
I would like to see a proof of that.
Here is a graph of an vs. n:

In the above graph, you can see that there is a local maximum with a period
of about 5, but not exactly 5. Furthermore, there is a second-order period
that is apparent in which the local maxima get closer to one about every
50 iterations, but, again, not exactly 50.
Here is a graph of the apparent period of this function vs. n:

And here is a graph of the apparent second-order period that you can
see in the first graph.
