When setting up SQL replication, the default retry count for the replication agent job step is 2147483647, retrying every minute. This means that if replication fails, the failure notification for the SQL Agent job won't be sent for 4085 years.
The issue is discussed in this MSDN blog, along with the suggestion that the retry count be changed from 2147483647 to 60 - so we would only wait one hour before being alerted to the failure: http://blogs.msdn.com/b/repltalk/archive/2010/08/25/sql-replication-agent-will-retry-for-4085-years.aspx
However it seems to me that the Microsoft SQL team must have selected the default of 2147483647 for a reason, and changing it to a much lower value must therefore have some negative impact.
So the question is - what is the best way to be alerted when a subscription breaks? (In our scenario, we are only the subscriber - and another company is the publisher)