It appears, that the line,
_dataframe_client.py:145
'time': int(ts.value / precision_factor)

Is causing a time rounding error when precision_factor is 'ns'. It is likely to be insufficient precision in the double64 type to hold this many seconds (inc 9 decimal places). The rounding error is in the order of (<100ns). This is consistent with the 15 sig.figures of the double64.

The issue is not present when using the line without the division e.g.:
'time': ts.value

but only for precision_factor, None or 'ns'. Other precisions (unsurprisingly) create time out of bounds on the POST to influxdb.

This line was added at commit:
02ddf0e