In this paper, we present a worst-case methodology for estimating the attainable spectral efficiency over end-to-end paths across a Flex Grid over multicore fiber (MCF) optical network. This methodology accounts for physical link noise, as well as for the signal-to-noise ratio in the Add module (
S
N
R
T
X
) of spatial-division-multiplexing-enabled reconfigurable optical add and drop multiplexers (SDM-ROADMs), introducing a dominant noise contribution over that of their Bypass and Drop modules. The proposed methodology is subsequently used to quantify the benefits that probabilistic constellation shaping (PCS) can bring to Flex-Grid/MCF dynamic optical backbone networks, compared to using traditional polarization-multiplexed modulation formats. In a first step, insight is provided into the spectral efficiency attainable along the precomputed end-to-end paths in two reference backbone networks, either using PCS or traditional modulation formats. Moreover, in each one of these networks, two
S
N
R
T
X
values are identified: the
S
N
R
T
X
yielding the maximum average paths’ spectral efficiency, as well as an
S
N
R
T
X
that, although slightly degrading the average paths’ spectral efficiency (by 10%), would yet enable a cost-effective SDM-ROADM Add module implementation. Extensive simulations are conducted to analyze PCS offered load gains under 1% bandwidth blocking probability. Furthermore, the study lastly focuses on finding out whether lower fragmentation levels in Flex-Grid/MCF dynamic optical backbone networks can push PCS benefits even further.