causing at least a portion of a radio-based application to be executed at a first runtime environment launched at a radio-based application pipeline processing server of a cloud provider network, wherein the radio-based application comprises a plurality of layers including a centralized unit (CU) layer, a distributed unit (DU) layer and a radio unit (RU) layer, wherein the portion executed in the first runtime environment comprises a first version of a workload of the radio-based application that implements functionality to (a) process inter-layer messages between a first layer of the plurality of layers and a second layer of the plurality of layers and (b) maintain state information about the inter-layer messages, wherein the radio-based application pipeline processing server comprises a network function accelerator card at which one or more network functions of the radio-based application are executed, and wherein the radio-based application pipeline processing server is located at a premise external to a data center of the cloud provider network; and

in response to determining that the workload of the radio-based application is to be executed at a second runtime environment,

transferring at least a subset of the state information of the workload of the radio-based application from the first runtime environment to the second runtime environment, without pausing the workload of the radio-based application; and

causing the workload of the radio-based application to be executed at the second runtime environment after the subset of the state information has been transferred.