On Thu, Jul 30, 2020 at 11:52 AM PG Bug reporting form
<noreply@postgresql.org> wrote:
> One field being updated is a timestamp provided as UTC text representation
> (e.g. '2020-07-29T22:30:00.124248Z') but stored as timestamp with time
> zone. The timestamp sub-second component is not consistently written -
> sometimes it is stored correctly, sometime it is stored incorrectly. Always
> the sub second part of the time (including more significant digits) and
> never the date/time from seconds upwards.
Kia ora,
Just to rule out another theory, if you run pg_controldata -D pgdata,
you can see which storage format is used for timestamps:
Date/time type storage: 64-bit integers
Before release 10, it was possible for it to use floating point
storage instead of integers; I wonder if that could be a factor here.
There's a note about that here:
https://www.postgresql.org/docs/9.5/datatype-datetime.html