On Wed, Sep 15, 2021 at 7:18 AM Marko Tiikkaja <ma...@joh.to> wrote: > So I've been looking at issues we used to have in production some time > ago which eventually lead us to migrating away from partial indexes in > some cases. In the end, I'm surprised how easy this (or at least a > similar case) was to reproduce.
> (I've tested this on 9.6, v11 and v13. 13 seems to be a bit better > here, but not "fixed", I think.) What about v14? There were significant changes to the microvacuum/index deletion stuff in that release: https://www.postgresql.org/docs/14/btree-implementation.html#BTREE-DELETION -- Peter Geoghegan
