Basically as title,

in the entrytmp table, ids are computed by adding the current usecs to the entry date. this is not guaranteed to be unique.

when adding to the entry table from the entrytmp table, ids are computed from the max id of the entrytmp table, and incremented. On two runs with the same max(id) - count(*) expression, you will have an id collision.

The entrytmp table seems to be used to get a date-oriented ordering for the id column. Why is this necessary? Why not order by the date columne in listWhereRaw() and sqlListWhere()?