SnapShot Magic

[Bob]

Hi:

Last week I installed all the bits and pieces to do automatic snapshots, and 
allow regular users to retreive lost/corrupted  data from hours, days, and 
weeks ago. It was all quite simple, and logical.

Now that it has been running for several days I have a few questions.

How can this be?:

$ snapshot list /usr
Filesystem          User   User%     Snap   Snap%  Snapshot
/usr             11399MB   37.1%    116MB    0.4%  daily.0
/usr             11399MB   37.1%    136MB    0.4%  daily.1
/usr             11399MB   37.1%     25MB    0.1%  hourly.0
/usr             11399MB   37.1%     28MB    0.1%  hourly.1
/usr             11399MB   37.1%    117MB    0.4%  hourly.2
/usr             11399MB   37.1%    120MB    0.4%  hourly.3

How can /usr:hourly.0 be 25MB, and  /usr:hourly.3 be 120MB ???

What's even more mystifying is this:

$ ls -al /usr/.snap
total 558260
drwxrwxr-x   2 root  operator          512 Sep 10 12:00 .
drwxr-xr-x  20 root  wheel             512 Aug 23 13:55 ..
-r--------   1 root  operator  33282639248 Sep 10 14:15 daily.0
-r--------   1 root  operator  33282639248 Sep 10 14:03 daily.1
-r--------   1 root  operator  33282639248 Sep 10 14:15 hourly.0
-r--------   1 root  operator  33282639248 Sep 10 14:15 hourly.1
-r--------   1 root  operator  33282639248 Sep 10 14:15 hourly.2
-r--------   1 root  operator  33282639248 Sep 10 14:03 hourly.3

/usr is a 33GB HW/Raid  partition, how can it possibly hold 6 33GB 
snapshots???

If I mount each one of these, they all report to df that they are indeed 33BB 
file systems!

How is this magic achieved? 

Is there a doc somewhere with an explaination?

TIA
Bob