and successfully read the first entries, but then there was a problem. I
then found in the source code (TermInfosWriter), that SkipDelta
is sometimes omitted. After fixing this problem, there apparently is still
another problem occurring after several hundred entries.
It looks like ProxDelta is missing too in these cases.
However I didn't find this in the source.
Therefore my question is whether there are exceptions from the scheme
given on the fileformats page: