Rule 2

A string that's intended to identify a UTC offset is never resolved in pg_timezone_names.abbrev.

You can discover, with ad hoc queries, that the string WEST occurs uniquely in pg_timezone_names.abbrev. Use the function occurrences() to confirm it thus:

with c as (select occurrences('WEST') as r)
select
  (c.r).names_name     ::text as "~names.name",
  (c.r).names_abbrev   ::text as "~names.abbrev",
  (c.r).abbrevs_abbrev ::text as "~abbrevs.abbrev"
from c;

This is the result:

 ~names.name | ~names.abbrev | ~abbrevs.abbrev
-------------+---------------+-----------------
 false       | true          | false

This means that the string WEST can be used as a probe, using the function legal_scopes_for_syntax_context()_:

select x from legal_scopes_for_syntax_context('WEST');

This is the result:

 WEST:               names_name: false / names_abbrev: true / abbrevs_abbrev: false
 ------------------------------------------------------------------------------------------
 set timezone = 'WEST';                                       > invalid_parameter_value
 select timezone('WEST', '2021-06-07 12:00:00');              > invalid_parameter_value
 select '2021-06-07 12:00:00 WEST'::timestamptz;              > invalid_datetime_format

You can copy-and-paste each offending statement to see each error occurring "live".

This outcome supports the formulation of the rule that this page addresses.

The values in pg_timezone_names.abbrev are useful only for decorating the to_char() rendition of a timestamptz value, thus:

set timezone = 'Atlantic/Faeroe';
with v as (
  select
    '2021-01-01 12:00:00 UTC'::timestamptz as t1,
    '2021-07-01 12:00:00 UTC'::timestamptz as t2
  )
select
  to_char(t1, 'hh24:mi:ss TZ TZH:TZM') as t1,
  to_char(t2, 'hh24:mi:ss TZ TZH:TZM') as t2
from v;

This is the result:

         t1          |          t2
---------------------+----------------------
 12:00:00 WET +00:00 | 13:00:00 WEST +01:00

The strings WET and WEST are respectively the Winter Time and Summer Time abbreviations for the Atlantic/Faeroe timezone.

Note: the strings in the pg_timezone_names.abbrev column don't necessarily uniquely denote a utc_offset value. You should think of these abbreviations, therefore, not as globally understood mappings. Suppose that you live in Los Angeles and your friend lives in Manila, and that you're planning a Holiday phone call. Imagine the confusion that would ensue if you said "Let's schedule our call for Saturday at seven PM PST"— (as people often do). Maybe you even speak that into a modern artificially "intelligent" assistant for your calendar app.

Try this:

drop function if exists appointment_details() cascade;
create function appointment_details()
  returns table (z text)
  language plpgsql
as $body$
declare
  set_timezone constant text  not null := $$set timezone = '%s'$$;
  tz_on_entry  constant text not null := current_setting('timezone');
  t constant   timestamptz   not null := '2021-12-25 19:00 America/Los_Angeles'::timestamptz;
begin
   execute format(set_timezone, 'America/Los_Angeles');
   z := 'Los Angeles time: '||to_char(t, 'Dy dd hh24:mi TZ'); return next;


   execute format(set_timezone, 'Asia/Manila');
   z := 'Manila time:      '||to_char(t, 'Dy dd hh24:mi TZ'); return next;

   execute format(set_timezone, tz_on_entry);
end;
$body$;

select z from appointment_details();

This is the result:

 Los Angeles time: Sat 25 19:00 PST
 Manila time:      Sun 26 11:00 PST

The calendar assistant, at least if it knew already what your location and your friend's location were and could map from these to timezones, would have to ask you "Do you mean PST in Los Angeles or PST in Manila on that date?".

A timezone abbreviation (in the world of human convention) is unique only within the scope of the timezone to which it belongs. Even then, it's not a very useful notion. with one caveat[1], because the combination of date-time and timezone tells you unambiguously if it's Winter Time or Summer Time.

This is why pg_timezone_names.abbrev is never used to resolve a string that's intended to identify a UTC offset.

[1] The caveat is that the abbreviation is useful, just once per year, during the period of one hour at the "fall back" moment when your wall-clock reads the same time after that moment as it did before. See the discussion, in the section "The plain timestamp and timestamptz data types" of this code example:

set timezone = 'America/Los_Angeles';
select
  to_char('2021-11-07 08:30:00 UTC'::timestamptz, 'hh24:mi:ss TZ (OF)') as "1st 1:30",
  to_char('2021-11-07 09:30:00 UTC'::timestamptz, 'hh24:mi:ss TZ (OF)') as "2nd 1:30";