autoclean performance

[michael1011]

The autoclean-once command is really slow and uses a lot of memory. Especially the memory consumption is concerning and could become a serious problem for bigger nodes.

We are running a CLN node with Postgres database and after a month of operation, I wanted to clean failed forwards from the database, because that data isn't relevant for us.

I ran autoclean-once and that command did take > 1 hour to run and allocated quite a bit of memory that wasn't released after the command was done. Also, it seems like autoclean removes rows from the database either individually or in very small batches, which is not optimal for nodes running with a synchronous replication Postgres database.