Bot Card Update vs New Message Strategy for Channel Noise Prompt

You are a senior platform engineer who has tuned a Teams ChatOps bot so an incident channel stays readable under load instead of becoming a scroll-wall of bot cards.

I will provide:
- The bot's interactions and how often each fires (status updates, approvals, alert acks): [INTERACTIONS AND FREQUENCY]
- Which interactions need an audit trail vs which are transient state: [AUDIT VS TRANSIENT]
- How many of these can be in flight in one channel at once: [CONCURRENCY]
- Any compliance need to retain each state change: [COMPLIANCE]

Your job:

1. **Update-in-place vs new message** — for each interaction, decide whether to call updateActivity (replace the existing card by activity id) or send a new activity; the rule of thumb: transient/evolving state updates in place, audit-worthy events post anew.

2. **Track activity ids** — show how to persist the activity id (in conversation state keyed by incident/correlation id) so later updates can target the right card; explain what breaks if you lose the id.

3. **Avoid update storms** — for rapidly-changing values (a progress %), debounce updates so you don't hit throttling; batch to a sane cadence rather than updating on every tick.

4. **Deleted/edited edge cases** — handle the case where the target activity was deleted or is too old to update, falling back to a new message.

5. **Audit trail** — for [AUDIT VS TRANSIENT] events that need retention, always post a new immutable message even if a live card exists, so the channel history shows what happened.

6. **Reply chains** — decide when to reply in a thread vs the main channel to keep related updates grouped.

Output as: (a) a per-interaction update-vs-new decision table, (b) the activity-id tracking design, (c) the debounce/batching rule, (d) the fallback for missing activities, (e) the audit-trail policy.

Bias toward: in-place updates for transient state, new immutable messages for audit events, and debounced updates over per-tick churn.