chore(mollifier): rewrite server-changes note for external readers

External changelog readers don't have context on internal phase numbering;
describe the feature itself (opt-in burst protection, default-off env vars,
shadow mode, dual-write activation) instead of "phase 1".

.server-changes/mollifier-burst-protection.md

@@ -0,0 +1,6 @@
+---
+area: webapp
+type: feature
+---
+
+Add the trigger mollifier: an opt-in burst-protection layer for the trigger hot path that detects per-env trigger storms and (when enabled) buffers them into Redis so the run engine can drain them at a sustainable rate. All new env vars default off, so existing deployments see no behaviour change. Operators can enable shadow-mode-only observability with `MOLLIFIER_SHADOW_MODE=1` (logs `mollifier.would_mollify` when an env exceeds the configured threshold, no buffer writes). Enabling `MOLLIFIER_ENABLED=1` with a per-org `mollifierEnabled` flag turns on dual-write monitoring: each over-threshold trigger is recorded in a Redis buffer alongside the normal `engine.trigger` call, and a background drainer pops and acks entries. Emits the `mollifier.decisions` OTel counter for per-env rate visibility.

.server-changes/mollifier-phase-1.md

@@ -1,6 +0,0 @@
----
-area: webapp
-type: feature
----
-
-Add the trigger mollifier (phase 1 — dual-write monitoring + shadow mode). New env vars (all default off), `evaluateGate` wired into the trigger hot path, lazy singletons for the dedicated mollifier Redis client and drainer. With `MOLLIFIER_SHADOW_MODE=1`, each trigger evaluates the per-env sliding-window rate counter and logs bursts as `mollifier.would_mollify` (no buffer write). With `MOLLIFIER_ENABLED=1` plus a per-org `mollifierEnabled` flag, the buffer is dual-written alongside `engine.trigger` and the no-op drainer pops/acks the entries. Emits the `mollifier.decisions` OTel counter. Behaviour with `MOLLIFIER_ENABLED=0` (default) is unchanged.