The strong‐motion downhole array (SMDA) in Taipei basin is examined, and the data quality, glitches, and systemic errors in its data are discussed. This seismic network is an array of arrays: the SMDA comprises a total of 32 triggered strong‐motion acceleration seismometers spanning eight sites. Each site has one seismometer at the surface and an additional two to four seismometers each collocated at the individual boreholes. Polarity reversals, swapped components, clock desynchronization, and bad components (flat‐line signal or aseismic noise) have all been observed and are shown to be occasional issues. Signal‐to‐noise ratios are generally excellent. The lack of known orientations at depth is the primary issue regarding data quality of the SMDA. Orientations of each borehole seismometer were recorded at the time of installation, but those initial values are not reliable for subsequent events. Furthermore, redetermined azimuthal orientations are inconsistent from event to event, suggesting that the borehole seismometer orientations change over time. Orientation wander is not associated with instrument maintenance. An iterative method to reliably determine borehole seismometer orientations is introduced. Data from each borehole seismometer are rotated and compared with that from a collocated reference station on the surface with known orientation until a maximum correlation is reached. This method is reliable for most events, but may become unreliable for local events in which the incidence of incoming seismic waves is near vertical and shorter wavelengths introduce complicated wave propagation within Taipei basin. Temporal analysis of borehole seismometer orientations shows not only drifting of azimuthal orientations over time, but also erratic changes in orientation in multiples of 90°. This behavior is suggestive of frequent polarity reversals and/or swapped components on one or both of the two horizontal components of each borehole seismometer.