Document slow-powermeter sensitivity for EFFICIENCY_SATURATION_THRESHOLD

The time-weighted saturation EMA accumulates faster per sample when
the powermeter update interval is large (e.g. >10 s), which can
cause unnecessary forced swaps.  Note the workaround (raise the
threshold) in both README.md and config.ini.example.

https://claude.ai/code/session_01AaF4EqZPib3pmM44w8DJXD

Author: claude

README.md

@@ -254,7 +254,10 @@ mid-interval failures still trigger a swap.
   healthy deprioritized battery instead of waiting for the next timed rotation.
   During a probe, the probe timeout is the main "never ramps" control; this
   threshold still matters after the probe succeeds and for already-active
-  batteries. Set to 0 to disable.
+  batteries. Set to 0 to disable. The saturation EMA is time-weighted, so
+  batteries with slower powermeters (>10 s update interval) accumulate saturation
+  faster per sample — if you see unnecessary swaps with a slow powermeter,
+  raise this value (e.g. to 0.8).
 - **SATURATION_DETECTION** (default true) — Track how well each battery follows
   its target. When a battery cannot deliver (full or empty), its share is
   reduced and redistributed to others.

config.ini.example

@@ -102,7 +102,9 @@ THROTTLE_INTERVAL = 0
 ## Force-rotate when a battery can't follow its target (e.g. externally limited
 ## to 0 W or fully discharged). After a probe succeeds, this still swaps out an
 ## active battery within a few seconds instead of waiting for the full rotation
-## interval. 0 disables.
+## interval. 0 disables. The saturation EMA is time-weighted, so slow
+## powermeters (>10 s updates) accumulate saturation faster per sample; raise
+## this value (e.g. 0.8) if you see unnecessary swaps with a slow meter.
 #EFFICIENCY_SATURATION_THRESHOLD = 0.4
 ## How quickly a swapped-out battery becomes eligible again. Applied each cycle
 ## while the battery has no target. 1.0 = never recover, lower = faster recovery.