A contract the size of JEDI can (and probably is/will) build in that kind of independence as a requirement. Which isn't to say "Amazon must build and maintain an OpenStack implementation for DoD", but rather, that Amazon must turn over all APIs and source code used to build the implementation. Obviously, source code and APIs isn't enough to keep everything running - there still needs to be operational expertise within organizational memory - but the DoD would be empowered to start to take over operations and plan a migration to some other company maintaining the cloud infrastructure, if it were really untenable to renew the contract in a decade.
Put it this way: there's an old aphorism about how owing the bank $1 million is your personal problem, owning the bank $1 billion is your bank's problem, and owing the bank $1 trillion is your country's problem. There would be a parallel for cloud infrastructure: if you have a $1 million cloud bill then you have a (vendor lock-in) problem; if you have a $500 million cloud bill then your vendor has a problem (because you represent a huge chunk of business that can evaporate if you move to on-prem); if you have a $1 billion cloud bill then the wider market would have a problem (because of the reverberating shock effects if your migration to cheaper infrastructure is unsuccessful).
It's the textbook definition of something which is too big to fail.