High statistics data sets from experiments at the Relativistic Heavy Ion Collider (RHIC) and the Large Hadron Collider (LHC) with small and large collision species have enabled a wealth of new flow measurements, including the event-by-event correlation between observables. One exciting such observable $\rho(v^{2}_{n},[p_{T}])$ gauges the correlation between the mean transverse momentum of particles in an event and the various flow coefficients ($v_n$) in the same event [1]. Recently it has been proposed that very low multiplicity events may be sensitive to initial-state glasma correlations [2] rather than flow-related dynamics. We find utilizing the IP-JAZMA framework that the color domain explanation for the glasma results are incomplete. We then explore predictions from PYTHIA-8, and the version for including nuclear collisions called PYTHIA-ANGANTYR, which have only non-flow correlations and the AMPT model which has both non-flow and flow-type correlations. We find that PYTHIA-ANGANTYR has non-flow contributions to $\rho(v^{2}_{n},[p_{T}])$ in p+O, p+Pb, O+O collisions that are positive at low multiplicity and comparable to the glasma correlations. It is striking that in PYTHIA-8 in p+p collisions there is actually a sign-change from positive to negative $\rho(v^{2}_{n},[p_{T}])$ as a function of multiplicity. The AMPT results match the experimental data general trends in Pb+Pb collisions at the LHC, except at low multiplicity where AMPT has the opposite sign. In p+Pb collisions, AMPT has the opposite sign from experimental data and we explore this within the context of parton geometry. Predictions for p+O, O+O, and Xe+Xe are also presented.