Sounds like they are betting on a few fast pipelines instead of several robust ones. With pair programming the speed of mutation is also halved which means the benefit of feature branches decreases (i.e. it is actually statistically less likely that the CI fails per unit of time).
I usually pair program when the work benefits from two noggins working on a problem. I.e. team up when the situation needs it. I've never understood forced pair programming. Information diffusion mechanisms which lack archivability and discoverability (i.e. wikis) sound fragile to me. As for accountability ... source control submits are fairly discoverable.
Intuitively, this sounds analogous to a situation of fast prototyping versus working on an established production code base - in the former you want to progress and iterate fast, in the latter you want to progress profitably by being damn careful.
Comments
Sounds like they are betting on a few fast pipelines instead of several robust ones. With pair programming the speed of mutation is also halved which means the benefit of feature branches decreases (i.e. it is actually statistically less likely that the CI fails per unit of time).
I usually pair program when the work benefits from two noggins working on a problem. I.e. team up when the situation needs it. I've never understood forced pair programming. Information diffusion mechanisms which lack archivability and discoverability (i.e. wikis) sound fragile to me. As for accountability ... source control submits are fairly discoverable.
Intuitively, this sounds analogous to a situation of fast prototyping versus working on an established production code base - in the former you want to progress and iterate fast, in the latter you want to progress profitably by being damn careful.