Very funny at the timing: There is a place in the PowerDNS daemon that stumbles over exactly this. At one point the C expression 1< <31-1 is calculated instead of (1< <31) -1. Result: since Saturday noon all PowerDNS installations have been running at maximum load, because the wait time between database queries is no longer being maintained and the database server is being bombarded with a massive number of database queries.
The problematic expression is in pdns/communicator.hh in the earliest method, which finds the domain that most urgently needs to be forwarded to the slaves, or most urgently requires an update from the master, depending on whether the server functions as master or slave or both. Details can be found in the archive of the pdns-dev mailing list. Here is the original article.