I report on the cumulative results from a program started 24 years ago designed to measure the orbital period change of recurrent novae (RNe) across an eruption. The goal is to use the orbital period change to measure the mass ejected during each eruption as the key part of trying to measure whether the RNe white dwarfs are gaining or losing mass over an entire eruption cycle, and hence whether they can be progenitors for Type Ia supernovae. This program has now been completed for two eclipsing RNe: CI Aquilae ( eruptions. The eclipse times during the tails of eruptions are systematically and substantially shifted with respect to the ephemerides from the eclipses in quiescence, with this being caused by shifts of the center of light during the eruption. These eclipse times are plotted on an O − C diagram and fitted to models with a steady period change (Ṗ ) between eruptions (caused by, for example, conservative mass transfer) plus an abrupt period change (ΔP ) at the time of eruption. The primary uncertainty arises from the correlation between ΔP withṖ , such that a more negativeṖ makes for a more positive ΔP . For CI Aql, the best fit is ΔP = −3.7 +9.2 −7.3 × 10 −7 . For U Sco, the best fit is ΔP = (+43 ± 69) × 10 −7 days. These period changes can directly give a dynamical measure of the mass ejected (M ejecta ) during each eruption with negligible sensitivity to the stellar masses and no uncertainty from distances. For CI Aql, the 1σ upper limit is M ejecta < 10 × 10 −7 M . For U Sco, I derive M ejecta = (43 ± 67) × 10 −7 M .