[al_james]

We use pg_repack and hit that exact problem (left over triggers etc). In the end we resorted to creating and dropping the pg_repack extension after each run.

(edit: Missed an "and")

[kawsper]

Do you have any idea why their docs are not mentioning this, or even why their software aren't doing that itself?

[cookiecaper]

I haven't used pg_repack in production, but on my development box with PgSQL 10.x, if you end the process early, pg_repack will complain on the next run and ask you to clean up by dropping and re-adding the extension. Maybe this occurs in the wild if the cleanup code is not reached, for whatever reason.