Bump era5_batch parallelism to 12 concurrent (CDS-Beta ceiling)

Raise local_limit 2→4 (12 concurrent cluster-wide) and rate_limit 10→30
per hour. Each job makes 2 serial CDS requests, so active workers map 1:1
to CDS slots; 12 sits safely under CDS-Beta's ~16 per-user ceiling while
the rate gate stays above steady-state so it stops being the bottleneck.

config/runtime.exs

@@ -168,11 +168,17 @@ if config_env() == :prod do
       iemre: 3,
       era5: 2,
       # No global_limit: the Pro tracker leaked phantom slots across deploys
-      # and stalled the whole backfill. The rate_limit is enough to protect
-      # CDS and lets the per-pod local_limit scale naturally with replicas.
+      # and stalled the whole backfill. The rate_limit protects CDS and the
+      # per-pod local_limit scales naturally with replicas.
+      #
+      # Parallelism sizing: each job makes 2 CDS requests *serially* (single-
+      # level then pressure-level), so each active worker holds 1 CDS slot at
+      # a time. CDS-Beta's per-user ceiling is ~16 concurrent. local_limit 4
+      # × 3 pods = 12 concurrent, leaving headroom. rate_limit 30/hour stays
+      # above steady-state throughput so it doesn't re-become the bottleneck.
       era5_batch: [
-        local_limit: 2,
-        rate_limit: [allowed: 10, period: {1, :hour}]
+        local_limit: 4,
+        rate_limit: [allowed: 30, period: {1, :hour}]
       ],
       rtma: 2,
       backfill_enqueue: 1,

lib/microwaveprop/workers/era5_month_batch_worker.ex

@@ -18,9 +18,9 @@ defmodule Microwaveprop.Workers.Era5MonthBatchWorker do
   starve the fast `Era5FetchWorker` router that also lives in the ERA5
   namespace.
   """
-  # The :era5_batch queue is rate-limited to 10/hour so a single failed run
-  # eats a meaningful chunk of throughput. Retrying 5×, combined with the
-  # generous backoff below, gives CDS up to a day to come back without
+  # The :era5_batch queue is rate-limited (see config/runtime.exs) to stay
+  # under CDS-Beta's per-user concurrent ceiling. Retrying 5×, combined with
+  # the generous backoff below, gives CDS up to a day to come back without
   # discarding the tile and silently dropping the historical contact.
   use Oban.Worker,
     queue: :era5_batch,