pull down to refresh

I have experienced 1-2 instances of uncompacted databases making HTLC times go through the roof. It usually goes like this:
  • A node takes 30s+ to respond to an HTLC, making the payment stuck for a good while
  • The node has a database that has not been compacted for a while
  • After compaction, speeds reduce to normal 2-3s over Tor.
With lots of Tor only nodes and no automatic database compaction, especially on Umbrel, I am partially sure this may be contributing to the overall slow payment problem.