Fragmented populations possess an intriguing duplicity: even if subpopulations are reliably extinction-prone, asynchrony in local extinctions and recolonizations makes global persistence possible. Migration is a double-edged sword in such cases: too little migration prevents recolonization of extinct patches, whereas too much synchronizes subpopulations, raising the likelihood of global extinction. Both edges of this proverbial sword have been explored by manipulating the rate of migration within experimental populations. However, few experiments have examined how the evolutionary ecology of fragmented populations depends on the pattern of migration. Here, we show that the migration pattern affects both coexistence and evolution within a community of bacterial hosts (Escherichia coli) and viral pathogens (T4 coliphage) distributed across a large network of subpopulations. In particular, different patterns of migration select for distinct pathogen strategies, which we term 'rapacious' and 'prudent'. These strategies define a 'tragedy of the commons': rapacious phage displace prudent variants for shared host resources, but prudent phage are more productive when alone. We find that prudent phage dominate when migration is spatially restricted, while rapacious phage evolve under unrestricted migration. Thus, migration pattern alone can determine whether a de novo tragedy of the commons is resolved in favour of restraint.