Whenever a project relies upon other projects to provide key inputs, so called dependencies, the risk goes up. When the dependency is a clear one it is usually well managed. The dependent project has a very keen interest in it being so. But where the dependency is more opaque, or an interim solution is required, things can get overlooked or a complacent attitude prevails. In both cases the problems are discovered late and the dependent project is thrown into a panic.
I'm currently working on a project that is part of a much larger program, with numerous work streams involving infrastructure, networking, database design, and application development. One of the key activities involves transmitting a large amount of data for testing purposes from a mainframe site to server farm for database development, I'm managing the database part. The files are large and in production will be carried over a dedicated network. The network has been delayed and won't be available for the early part of the development effort. So an alternative transmission connection was required. This is when the chances of the wheels coming off started to increase, because once the network delay was identified and the question of transmission was raised the answer was; don't worry we'll move it using a virtual private network (VPN), no problem.
The moment this glib statement was made then this cynical PM immediately went on red alert. Why? Because passed experience with this company warned me that they were grossly underestimating the task and ignoring firewall permissions, server builds and authorities, and bandwidth. My tech team where equally blasé, relying on the fact that they had stipulated that they need the files by X date and that it was other peoples job to deliver them. I didn't share this sanguine attitude. I started badgering everyone in the whole transmission chain to give me the details of who was doing what, and when it would be done. Needless to say they all had other more pressing problems and without my push we'd never had made the date. Even so it was touch and go.
Interfaces between projects and programs are always a problem. You have to identify them, even if they are not obvious, then interlock with their managers, and constantly manage that interlock. People can forget their commitments: don't let them.