Carbon dioxide (CO<sub>2</sub>) is one of the most important anthropogenic greenhouse gases in the atmosphere, and accurate monitoring of CO<sub>2</sub> concentration is essential for carbon flux inversion and emission reduction policies. Spaceborne integrated path differential absorption (IPDA) lidar enables day-and-night observations, high-latitude coverage, and reduced sensitivity to clouds and aerosols, showing great potential for monitoring the column-averaged dry-air mole fraction of CO<sub>2</sub> (XCO<sub>2</sub>). However, single-shot XCO<sub>2</sub> retrievals from spaceborne IPDA lidar generally have relatively large random errors, and their along-track averaging is affected by surface conditions and data gaps. Based on 2023 DQ-1 spaceborne IPDA lidar observations, this study calculated global single-shot XCO<sub>2</sub> random errors using the return-signal SNRs at the online and offline wavelengths. The results show pronounced spatial heterogeneity in single-shot random errors. Larger errors occur over water bodies, permanent snow and ice, and some complex land cover types, whereas day–night differences are relatively weak. Combining MCD12C1 land cover data with random-error statistics, the global observation scenes were classified into four surface random-error categories. Allan variance analysis of continuous observation segments suggests averaging approximately 160, 200, 232, and 296 observations for the four categories, corresponding to typical along-track scales of about 54, 67, 77, and 100 km. To address data gaps and along-track discontinuities, fixed-spatial-resolution and random-error-threshold-controlled averaging strategies were compared. The fixed-spatial-resolution strategy maintains more stable spatial representativeness and is more suitable as the default averaging option, whereas the threshold-controlled strategy improves error consistency but may substantially expand the averaging window. For land–sea boundaries and other rapid surface-transition regions, a boundary mixed dynamic averaging strategy was further proposed, with surface-category fractions recommended as quality indicators. These results support random-error evaluation and averaging-strategy design for spaceborne IPDA lidar XCO<sub>2</sub> products.