I've never run a Lemmy or kbin instance, but have run a mastodon instance and am familiar with distributed systems in general.
Probably both the source instance and the destination instance have a queue of federated messages to process and the message needs to be sent to the destination and processed into the destination database respectively before it'll show up on the destination of the worker threads get overwhelmed.
I'd imagine that recently with the influx of users, activity, and the size of the federation, a number of instances probably sometimes have some pretty deep queues to work through at times.
Ten hours isn't ideal and hopefully isn't common, but doesn't seem out of the question in the case of connectivity issues or spikes im activity.