Inordinate number of "missing segment" errors from check_seismic followup

We've been seeing a lot of analysis errors recently, at both sites but particularly at LLO. It seems, though, that it's always the same failure in the seismic.check_seismic function:

2019-11-12 12:19:10,104 [analyze.analyze_event] exception in check_seismic:
Traceback (most recent call last):
  File "/home/lockloss/src/locklost/locklost/analyze.py", line 53, in analyze_event
    func(event)
  File "locklost/plugins/seismic.py", line 26, in check_seismic
    seismic_channels = data.fetch(config.SEISMIC_CHANNELS, segment)
  File "locklost/data.py", line 165, in fetch
    return func(channels, start, stop)
  File "locklost/data.py", line 148, in frame_fetch_gwpy
    data = gwpy.timeseries.TimeSeriesDict.find(channels, start, stop, frametype=config.IFO+'_R')
  File "/usr/lib/python2.7/site-packages/gwpy/timeseries/core.py", line 1284, in find
    cache = io_datafind.find_urls(observatory, frametype, start, end)
  File "/usr/lib/python2.7/site-packages/gwpy/io/datafind.py", line 333, in wrapped
    return func(*args, **kwargs)
  File "/usr/lib/python2.7/site-packages/gwpy/io/datafind.py", line 641, in find_urls
    on_gaps=on_gaps)
  File "/usr/lib/python2.7/site-packages/gwdatafind/http.py", line 433, in find_urls
    raise RuntimeError(msg)
RuntimeError: Missing segments: 
[1257617600 ... 1257617641)

I suspect that there's a segment start/end calculation error somewhere.