Consistency and precision of very long baseline interferometry source coordinate time series
Coordinate time series of extragalactic radio sources from very long baseline interferometry observations are widely used to evaluate source positional stability, select stable sources for celestial reference frames, assess reference-frame axis stability, and investigate source-related astrometric variability. Their external consistency, realistic errors, and relation to processing strategies and analysis configurations remain to be investigated. We aim to compare coordinate time series solutions from different analysis centers and investigate how processing strategies and configurations contribute to their differences. We collected eight solutions from seven analysis centers and selected 496 common sources after data selection, reference-frame alignment, and extraction of common observing sessions. Inter-solution differences were characterized with pairwise positional offsets and correlation analysis, and realistic positional errors with the N-cornered-hat (NCH) method and bootstrap resampling. Most solutions have median weighted root mean square (WRMS) values of a few hundred microarcseconds $(\mathrm{μas})$, and the median NCH-derived precision is about 200-300 $\mathrm{μas}$ in right ascension and 250-400 $\mathrm{μas}$ in declination. Solutions with similar strategies, software, or source constraints show higher consistency, with more similar WRMS values, stronger correlations, and closer precision estimates. Consistency is affected not only by the global or independent mode distinction but also by detailed analysis configuration. NCH-derived precision depends on declination, being poorer in the southern sky, and stabilizes with more observing sessions. Processing strategies and configurations introduce marginal differences; solutions remain generally consistent. Inter-solution comparisons independently estimate realistic stochastic errors beyond formal errors.