One reason for example was (which is now fixed in the next release update) a logic problem in block processing. This resulted in a CPU spike when processing a block while there are many transactions on the network at the same time. During that time where the CPU spiked to 100% the node couldn't do anything else, so he fall behind / starved. I.e. chances are high he lands on a fork.

...
As you all probably have assumed, we have encountered numerous issues with the delegate system, and the security of the rise-core code base. A couple of highlights:

• Slow system speed - 10> Transactions per second - Above this causes forks
• DDoS Vulnerabilities in multiple core system packages that require significant updates to resolve
• Spotty delegate communication
• High resource usage for nodes

The above issues are difficult to resolve in the current code base (As evidenced by the original developers lack of resolution up until this point).

In order to find a solution for this, we are going back to the drawing board, and designing an entirely new architecture for the system, not based on anything that is currently in existence. We will be writing a new codebase from scratch.
...