Let the file 'AAA' contains the following script:

recreate table tmp(id int not null primary key using index tmp_id_pk);
commit;
set transaction no wait isolation level read committed;
set term ^;
execute block as
begin
  insert into tmp values(1);
  insert into tmp values(2);
  in autonomous transaction do begin
     insert into tmp values(1);
     when any do begin
        --exception; -- <<<<<<<<<<<<<< ATTENTION: **NO** handle exception here
     end
  end
end^
set term ;^
commit;
select id from tmp;
select count(*) from tmp;
commit;

Create NEW EMPTY database, connect to it via ISQL -n and then:

SQL> IN AAA;

Result:

          ID
============
           1
           2
           1


       COUNT
============
           3

So, we have duplicate in PK. The same result will be if PK replaced by UK.
--------------------------------------
Moreover,
1) GFIX -V -FULL will *NOT* alert that table TMP has some problem in it's PK;
2) gbak -b will create UNRECOERABLE .fbk.
If then we'll try gbak -c test.fbk test_new.fdb then following messages will fills the screen:
...
...
gbak: The unique index has duplicate values or NULLs.
gbak: Delete or Update duplicate values or NULLs, and activate index with
gbak: ALTER INDEX "TMP_ID_PK" ACTIVE;
gbak:Index "TMP_ID_PK" failed to activate because:
gbak: The unique index has duplicate values or NULLs.
gbak: Delete or Update duplicate values or NULLs, and activate index with
gbak: ALTER INDEX "TMP_ID_PK" ACTIVE;
gbak:Index "TMP_ID_PK" failed to activate because:
gbak: The unique index has duplicate values or NULLs.
gbak: Delete or Update duplicate values or NULLs, and activate index with
gbak: ALTER INDEX "TMP_ID_PK" ACTIVE;

... etc ... (these messages repeats infinitely)