discard slices rather than reslicing them

In a previous version of this, I thoughtfully resliced slices
to [:0] rather than allocating. That's a great idea, except that
you don't necessarily ever see a given shard again. If you
don't see that shard again, or if you don't get as many items
in it, all the old record pointers are still sitting there in
reachable parts of the slice's backing store, preventing them
from ever being garbage collected.

import_manager.go

@@ -128,27 +128,29 @@ func recordImportWorker(id int, client *Client, field *Field, chans importWorker
 
 readRecords:
 	for recordBatch := range recordChan {
+		// It's fine to overrun our allowed batch size slightly, and
+		// we don't want to generate separate batches for part of a
+		// 16-item batch.
 		for _, record := range recordBatch {
 			recordCount++
 			shard := record.Shard(shardWidth)
 			if batchForShard[shard] == nil {
 				batchForShard[shard] = make([]Record, 0, batchSize)
 			}
 			batchForShard[shard] = append(batchForShard[shard], record)
-
-			if recordCount >= batchSize {
-				for shard, records := range batchForShard {
-					if len(records) == 0 {
-						continue
-					}
-					err = importRecords(id, client, field, shardNodes, shard, records, options, statusChan, state)
-					if err != nil {
-						break readRecords
-					}
-					batchForShard[shard] = batchForShard[shard][:0]
+		}
+		if recordCount >= batchSize {
+			for shard, records := range batchForShard {
+				if len(records) == 0 {
+					continue
+				}
+				err = importRecords(id, client, field, shardNodes, shard, records, options, statusChan, state)
+				if err != nil {
+					break readRecords
 				}
-				recordCount = 0
+				delete(batchForShard, shard)
 			}
+			recordCount = 0
 		}
 	}