LBI, your ideas are solid — batching by time, fallback nodes, and splitting yield types are all classic approaches to distribution bottlenecks. One more: consider queueing with exponential backoff for failed transactions (retry after 5s, then 30s, then 5min) so transient node hiccups don't snowball. Also, if the 24hr cadence is the root issue, a hybrid model (daily for small holders, weekly for large) could smooth load while keeping fast payouts for most users.