DNS clients that spend more time appending suffixes to ambiguous names and retrying their searches will take longer before giving up. This can cause significant slowdowns in applications that perform a lot of DNS queries.
It can also create a security concern if DNS clients erroneously resolve a name that is under the control of an external, malicious entity. Appending DNS suffixes is basically the opposite of devolution, which can present similar concerns. I'll copy the example from the Windows IT Pro website (which is primarily about devolution, but also applies somewhat to appending suffixes) :
A domain-joined computer's primary domain suffix is mycompany.fl.us
(mycompany is located in Florida, hence the extension fl.us) and tries
to connect to mailserver1. In this example, the DNS client will try to
resolve mailserver1.mycompany.fl.us and mailserver1.fl.us. The last
name in this list, mailserver1.fl.us, is outside of the control of my
company. If a malicious person has registered mailserver1.fl.us in the
DNS, the name resolution will succeed, the domain-joined computer will
try to connect to it, and the malicious user could spoof an internal
server.
So why would you want to turn it on? You might want to give DNS clients the added flexibility of hopefully being able to resolve ambiguous names. But it could theoretically lead to a security concern. So it's up to the administrator to decide what is more appropriate for his or her environment.
Further reading:
http://blogs.technet.com/b/networking/archive/2009/04/16/dns-client-name-resolution-behavior-in-windows-vista-vs-windows-xp.aspx
And:
http://windowsitpro.com/networking/whats-dns-name-devolution