Dense wavelength division multiplexing (DWDM) transport networks are evolving to support dynamic setup/teardown of wavelength services. This is becoming possible mainly via node architectures with flexible add/drop port utilization, where an already deployed transponder can be tuned to any wavelength and configured to transmit/receive via any output/input direction, and advanced control and management software, which provide the means to remotely reconfigure equipment end to end. In long-haul networks, exploiting these capabilities to quickly set up new services also demands that 3R regenerators (performing reamplification, reshaping, and retiming) have been deployed in advance. This paper proposes a framework that, first, characterizes stochastically the fitness of each network node as a placeholder for predeployed regenerators based on the network and traffic information available and, second, distributes the set of regenerators to match as closely as possible the fitness values. Network simulation on a reference long-haul network under dynamic traffic is used to evaluate the effectiveness of four fitness estimation strategies, each based on a different set of forecasts about the expected network operating conditions. The results highlight that when the main assumptions embedded in these strategies hold during network operation significant benefits can be attained, in terms of lower service blocking probability and/or reduced number of predeployed 3R regenerators, by exploiting additional information when computing the node's fitness values.