Submitted by: @hvlad

RECREATE TABLE T (ID INT NOT NULL, DESCR VARCHAR(32) NOT NULL);
COMMIT;

INSERT INTO T (ID, DESCR) VALUES (1, 'No F1 field');
COMMIT;

ALTER TABLE T ADD F1 VARCHAR(16) DEFAULT 'XYZ' NOT NULL;
COMMIT;

INSERT INTO T (ID, DESCR) VALUES (2, 'F1 field, default XYZ');
COMMIT;

SELECT * FROM T;
COMMIT;

ALTER TABLE T ALTER COLUMN F1 SET DEFAULT 'ABC';
COMMIT;

INSERT INTO T (ID, DESCR) VALUES (3, 'F1 field, default ABC');
COMMIT;

SELECT * FROM T;
COMMIT;

After the first select all is as expected:

      ID DESCR                            F1

============ ================================ ================
1 No F1 field XYZ
2 F1 field, default XYZ XYZ

after the second select
expected

      ID DESCR                            F1

============ ================================ ================
1 No F1 field ABC
2 F1 field, default XYZ XYZ
3 F1 field, default ABC ABC

actual

      ID DESCR                            F1

============ ================================ ================
1 No F1 field XYZ
2 F1 field, default XYZ XYZ
3 F1 field, default ABC ABC

Note value of the field F1 at the first record: it is expected that is should be the same as latest DEFAULT value.
Also note that 2nd and 3rd INSERTs assigns correct value to the omitted field - same as latest DEFAULT value.