+/*
+ * flow control notes
+ * to ensure articles go away eventually
+ * separate queue for each input file
+ * queue expiry
+ * every period, check head of backlog queue for expiry with SMretrieve
+ * if too old: discard, and check next article
+ * also check every backlog article as we read it
+ * flush expiry
+ * after too long in SEPARATED/DROPPING ie Separated/Finishing/Dropping
+ * one-off: eat queued articles from flushing and write them to defer
+ * one-off: connfail all connections which have any articles from flushing
+ * newly read articles from flushing go straight to defer
+ * this should take care of it and get us out of this state
+ * to avoid filling up ram needlessly
+ * input control
+ * limit number of queued articles for each ipf
+ * pause/resume inputfile tailing
+ */
+