Answer
Bandwidth deserves the most rigorous check because it is the constraint most likely to be underestimated. Initial seeding of an IBM i environment into a cloud target can involve moving a large volume of Db2 for i data and IFS content, and ongoing replication has to keep pace with daily change volume without saturating the link needed for normal production traffic. Ask the vendor for a sizing exercise based on actual journal volume and IFS growth, not a generic bandwidth recommendation.
Licensing assumptions are the other area that catches buyers off guard. IBM i software licensing in a cloud or hosted environment can carry different terms than an on-premises LPAR, and some ISV products license by processor tier or user count in ways that do not translate cleanly to cloud infrastructure. Confirm before signing that every application the business depends on is properly licensed to run at the cloud target, not just during a declared disaster but during test windows too, since untested licensing gaps tend to surface at the worst possible moment. Finally, get specific about who owns failover operations: a managed service that handles infrastructure but leaves application-level failover to an internal team that has never run PTFs or verified subsystem startup in that environment is not fully ready.