A workaround solution could be filter trimming. This will set the start and end point of a filter (such as mask or blur) without requiring a split. If there is a long video clip that needs five sections blurred, then five blur filters can be added, each with its own start/end point. The underlying video clip would not need to be split at all.
Having said that, the best solution is probably to convert audio files that are not seek-friendly (such as MP3) to a format that is (such as WAV), and ensure that no sample rate conversion is happening in the pipeline. As in, all input files and export settings agree to 48k instead of mixing and matching 48000 and 44100.