See this comment on their previous paper: https://arxiv.org/abs/2306.07628
The authors of the comment show that the "gravitational pair-production" rate used in the work in OP comes from truncating the covariant heat-kernel (proper-time) expansion of the one-loop effective action at second order in curvature, an approximation that is valid only in weak-field regions where all curvature invariants satisfy |R| · ℓ² ≪ 1 (where ℓ is the Compton wavelength). When that same expression is pushed into the high-curvature interior of a neutron star -- where the inequalities fail by many orders of magnitude -- the series is no longer asymptotic and its early terms generate a spurious imaginary part. Because the paper's entire mass-loss mechanism and lifetime bound follow from that uncontrolled imaginary term, its conclusions collapse.
Simply put, it doesn't even correspond to known experiments. It's entirely driven by a narrow artefact and has no physical basis.