>Yowsa! Maybe your problem was "crossing midnight" with "t" value (>86400) or was "t" now part of the new day (date+1 t%86400)?

no, it was like your problem looking at the data in VFP it was not obvious that the internal values were not what was displaying. I had to put the record values into a temp table and use my hex file viewer to see that the data was one bit smaller than the "equivalent" datetime constant stored in the same table.