I upgraded from about 1.4.1 + local patches that shouldn't be relevant (I added a "bulkimport" method to the mailgw, 
similar to mbox, but aware of message-ids so you can import the same message multiple times and it will be ignored if 
redundant).

Then I stopped the server and ran "roundup-admin -i `pwd` migrate" and "roundup-admin -i `pwd` reindex" in the 
instance directory.  I assume the problem is the reindex because migrate said nothing needed to be done.

After that all attempts to access the server gave a traceback as follows:

steve@tleeps1:/var/www/trackers/xemacs$ roundup-admin -i `pwd`
Roundup 1.4.8 ready for input.
Type "help" for help.
roundup> table priority
Traceback (most recent call last):
  File "/usr/local/bin/roundup-admin", line 6, in ?
    run()
  File "/usr/local/lib/python2.4/site-packages/roundup/scripts/roundup_admin.py", line 38, in run
    sys.exit(tool.main())
  File "/usr/local/lib/python2.4/site-packages/roundup/admin.py", line 1524, in main
    self.interactive()
  File "/usr/local/lib/python2.4/site-packages/roundup/admin.py", line 1462, in interactive
    self.run_command(args)
  File "/usr/local/lib/python2.4/site-packages/roundup/admin.py", line 1425, in run_command
    self.db = tracker.open('admin')
  File "/usr/local/lib/python2.4/site-packages/roundup/instance.py", line 131, in open
    db.post_init()
  File "/usr/local/lib/python2.4/site-packages/roundup/backends/rdbms_common.py", line 230, in post_init
    save = save | self.upgrade_db()
  File "/usr/local/lib/python2.4/site-packages/roundup/backends/rdbms_common.py", line 281, in upgrade_db
    self.fix_version_4_tables()
  File "/usr/local/lib/python2.4/site-packages/roundup/backends/rdbms_common.py", line 308, in fix_version_4_tables
    self.add_class_key_required_unique_constraint(cn, klass.key)
  File "/usr/local/lib/python2.4/site-packages/roundup/backends/rdbms_common.py", line 577, in
    add_class_key_required_unique_constraint
    self.sql(sql)
  File "/usr/local/lib/python2.4/site-packages/roundup/backends/rdbms_common.py", line 161, in sql
    self.cursor.execute(sql)
psycopg2.ProgrammingError: relation "_status_key_retired_idx" already exists

(similar tracebacks would result from attempts to access the web interface, this is nice and compact, though).

The workaround was to fire up psql, connect to the postgresql database, and execute "DROP INDEX 
_status_key_retired_idx;".  This needed to be run for most of the properties that an issue has.  My tracker is now running 
normally as far as I can tell.