Returns true when candidate should replace existing as the representative agent run
for a single ConversationDetailID. Lower tier wins (primary root run beats continuation
synthesis run, which beats child sub-agent run); within the same tier, a strictly newer run
wins (so user retries of a message replace the earlier run while DESC-ordered queries keep
the first row when timestamps are tied or omitted in test fixtures).
Returns true when
candidateshould replaceexistingas the representative agent run for a singleConversationDetailID. Lower tier wins (primary root run beats continuation synthesis run, which beats child sub-agent run); within the same tier, a strictly newer run wins (so user retries of a message replace the earlier run while DESC-ordered queries keep the first row when timestamps are tied or omitted in test fixtures).