It seems that a concurrent UPDATE can restart heap_lock_tuple() even if it'snot necessary. Is the attached proposal correct and worth applying?
-- Antonin HouskaWeb: https://www.cybertec-postgresql.com