Update the timestamp values in a column so that the latest timestamp is current time while maintaining the relative time difference between rows

I have a table in a Postgres database with a column whose values are timestamps for a period of time in the past. How can I update the values in that column so the highest value (the most recent timestamp) is the current timestamp, keeping the same interval between the other ones? For example, if the current values are:

2019-05-23 10:00:00
2019-05-23 10:30:00
2019-05-23 11:45:00

and I run the query at 2020-07-16 14:45:00, then I want the values to be updated to be:

2020-07-16 13:00:00
2020-07-16 13:30:00
2020-07-16 14:45:00

Answer

In order to update the timestamps in-place and without manually entering the current timestamp, the following query can be used:

UPDATE
   sample_metrics_data 
SET
   "timestamp" = sub.new_ts 
FROM
   (
      SELECT
         timestamp,
         (
            NOW() - (max(timestamp) over ()) + timestamp
         )
         AS new_ts 
      FROM
         sample_metrics_data 
   )
   sub 
WHERE
   sample_metrics_data.timestamp = sub.timestamp;

Subquery (or a CTE) is needed as UPDATE does not allow the use of window functions.