Both swerve and mecanum are equally agile when executed well.
Mecanum is easier to execute well and requires fewer resources ($s, mass,...) and less learning or insitutional knowledge to do so. The rollers, however, do compromise traction.
Swerve provides agility without compromising traction, but it is resource intensive. 1640 uses swerve and we are very pleased with the results, but this has come at a considerable investment in learning. We also have to deal with having a finished drive-train rather late in the build season due to long fabrication and assembly times. Control software is also non-trivial. So there is a price to pay for this traction.
It is clear that the investment and cost of swerve has imposed limitations on other systems and capabilities on our robots.
So, like all real design decisions, this one is about what is right for you and your situation. All good designs are good compromises.
More serve info is available at the following link to "swerve central":
http://wiki.team1640.com/index.php?title=Swerve_Central