update: As added context, please also see Paul Handy's (IOTA Developer) github issue:
False-Positive Subtangle Rejection
Large subtangles are currently skipped if they fall below a node's configured maximum depth (counted in milestone subtangles). I would suggest that while this reduces the amount of tying in of 'stale' subtangles, it is detrimental and ultimately goes against one of the main advantages of Tangle - that offline subtangles may merge in the future.
One of the features of IOTA is that sub-tangles can be constructed and then attached to to the main tangle in the future.
With IRI version 1.5.2, how does the prevention of lazy tip selection change the tangle's partition tolerance?
For Further Reference, please see these resources below
[IOTA solves] Partition Intolerance
Blockchain-based currencies are unable to survive long-sustained partitioning of the network because this may lead to reversal of a large number of transactions. It is also impossible to initiate an intentional partitioning in cases when it is required.
A quick-easy reference to it by @zauz here on iota.stackexchange:
A @medium article on it called Sharding the IOTA Tangle Effectively.
Indeed, this is even covered in an AMA here: