sync-v2 §10.7: mock-provider soak/load harness + poll jitter

- loadtest/soak_sync.py + mock_provider.py: fully isolated soak for acceptance
  criterion 7. Creates a throwaway webmail_soak DB + uses Redis DB 14 (both torn
  down), simulates N accounts (default 1000, 20% active sessions) across 3
  in-process coordinators with a mock MailProvider (no real IMAP). Asserts:
  all accounts owned, persistent sockets <= hot*1.05, oldest lag < POLL_COLD*1.15,
  RSS stable. Steady state = after every account has completed its first sync.
  Configurable via SOAK_* / POLL_* env for fast smoke runs.
- engine: poll_all_loop and secondary_poll_loop now jitter their waits (first at a
  random point in the interval, later +/-10%) so a fleet claimed together doesn't
  poll in lockstep — anti-thundering-herd, a real improvement surfaced by the soak.

Validated: 60-account smoke passes all four criteria; full gate green (238 tests).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

backend/app/sync/engine.py

@@ -5,6 +5,7 @@ from __future__ import annotations
 import asyncio
 import contextlib
 import logging
+import random
 import uuid
 from typing import Any
 
@@ -353,10 +354,15 @@ async def secondary_poll_loop(
     """
     stop = stop or asyncio.Event()
     period = interval if interval is not None else get_settings().poll_warm
+    first = True
     while not stop.is_set():
-        # Wait first: the initial state was already covered by the backfill.
+        # Wait first (initial state was covered by backfill). Spread the fleet: the
+        # first wait is a random point in the period, later waits jitter ±10%, so
+        # accounts claimed together don't poll in lockstep (thundering herd).
+        wait = random.uniform(0, period) if first else period * random.uniform(0.9, 1.1)
+        first = False
         with contextlib.suppress(TimeoutError):
-            await asyncio.wait_for(stop.wait(), timeout=period)
+            await asyncio.wait_for(stop.wait(), timeout=wait)
         if stop.is_set():
             break
         await _poll_secondary_once(account_id)
@@ -371,9 +377,14 @@ async def poll_all_loop(
     is held between passes. Correctness matches the hot tier — only latency differs.
     """
     stop = stop or asyncio.Event()
+    first = True
     while not stop.is_set():
+        # Spread the fleet: first poll at a random point in the interval, then
+        # jitter ±10% so a batch claimed together doesn't poll in lockstep.
+        wait = random.uniform(0, interval) if first else interval * random.uniform(0.9, 1.1)
+        first = False
         with contextlib.suppress(TimeoutError):
-            await asyncio.wait_for(stop.wait(), timeout=interval)
+            await asyncio.wait_for(stop.wait(), timeout=wait)
         if stop.is_set():
             break
         await poll_folders_once(account_id, include_inbox=True)