On Fri, Aug 22, 2025 at 08:45:11AM +0200, Claudio Jeker wrote:
> The switch to ibufq introduced a bug with the imsg_pending tracking.
> peer_imsg_flush() did not adjust imsg_pending so when a peer is reset
> in the wrong moment imsg_pending becomes off and so peer_work_pending
> would always return 1.
> 
> I dislike this imsg_pending tracking and decided it is better to
> just walk the peertable in peer_work_pending. This is more work but less
> error prone.

Agreed.

> This seems to fix my spinning RDE at 100% issue (I hope).

ok tb