Basicstats aggregator (1.9.0) failing to "drop_original"

[glinton]

Relevant telegraf.conf:

[agent]
  interval = "1s"
  round_interval = true
  metric_batch_size = 1000
  metric_buffer_limit = 10000
  collection_jitter = "0s"
  flush_interval = "1s"
  flush_jitter = "0s"
  omit_hostname = true

[[inputs.exec]]
  command = "echo -en '{\"verb\":\"GET\",\"request\":\"/time/to/awesome\", \"status\":200}'"
  data_format = "json"
  json_string_fields = ["request"]
  tag_keys = ["verb"]
  timeout = "1s"

[[aggregators.basicstats]]
	period = "5s"
	stats = ["count"]
	name_suffix = "_count"
	drop_original = true

[[outputs.file]]

System info:

Telegraf 1.8.3 and 1.9.0
Ubuntu 18.04

Steps to reproduce:

1. telegraf-1.9.0 --config telegraf.conf

Expected behavior:

exec_count,verb=GET status_count=6 1543938920000000000
exec_count,verb=GET status_count=5 1543938925000000000
exec_count,verb=GET status_count=5 1543938930000000000
exec_count,verb=GET status_count=5 1543938935000000000
exec_count,verb=GET status_count=5 1543938940000000000
exec_count,verb=GET status_count=5 1543938945000000000
exec_count,verb=GET status_count=5 1543938950000000000
exec_count,verb=GET status_count=5 1543938955000000000
exec_count,verb=GET status_count=5 1543938960000000000
exec_count,verb=GET status_count=5 1543938965000000000
exec_count,verb=GET status_count=5 1543938970000000000

Actual behavior:

exec_count,verb=GET status_count=5 1543938854000000000
exec_count,verb=GET status_count=5 1543938859000000000
exec_count,verb=GET status_count=5 1543938864000000000
exec_count,verb=GET status_count=5 1543938869000000000
exec_count,verb=GET status_count=5 1543938874000000000
exec,verb=GET request="/time/to/awesome",status=200 1543938874000000000
exec_count,verb=GET status_count=4 1543938879000000000
exec,verb=GET request="/time/to/awesome",status=200 1543938879000000000
exec_count,verb=GET status_count=4 1543938884000000000
exec,verb=GET request="/time/to/awesome",status=200 1543938884000000000

[glinton]

Previously, the metrics that were processed outside of the sample window (missed by a subsecond) were just dropped, 1.9.0 [https://github.com/influxdata/telegraf/blob/1.9.0/internal/models/running_aggregator.go#L126](doesn't drop) the metric, but instead adds it in. Upside is no metrics are lost, downside is it doesn't get aggregated into the next window and is "lost" if one only cares about the aggregated metrics.