There is a point of confusion currently in the System Center 2012 Configuration Manager product, and that revolves around fallback DP and how they get used.
The scenario is this. You setup a content boundary group associated with 2 distribution points. You put the content on all of them. You associate the boundary of a machine to that content boundary group. What most people think is that the machine will get content from either of the 2 DPs.
Normally, this would be the case. However, if some of those DP are marked as fallback DP, they will then be "bucketized" differently in the DP selection process. If the application in question is not enabled for fallback then that fallback DP will not be offered to the requesting client and all traffic will be limited to only the remaining DP.
Here is an example scenario to help illustrate the current behaviour:
"We have two DPs: DP1 and DP2.
DP1 has the option “Allow fallback source location for content” checked.
We have only one Boundary Group : BG1
BG1 has DP1 and DP2 as “assigned” site systems
Deployment type doesn’t have Fallback enabled (default option).
In that case, clients which ”belong” to the BG1, will not use DP1 at all. It won’t be “offered” by the MP."
Put another way... watch out what DPs you mark for fallback. The DP will either be a fallback DP, or a preferred DP.., but not necessarily both at the same time.