First batch of results. One observation does not a theory make, but if anyone has any test suggestions feel free to share.
Speedtests are performed daily at 00:05, 06:05, 11:05, 23:05. The idea is to compare the night time speeds to best-case day time speeds.
Code:
Columns are server, download, upload, ping, timestamp
| (Cape Town) speedtest.tb.mtn.co.za | 37.87 | 7.80 | 43.23 | 2017-10-25 23:05:28 |
| (Cape Town) speedtest.tb.mtn.co.za | 34.21 | 7.13 | 39.07 | 2017-10-26 00:05:27 |
| (Cape Town) speedtest.tb.mtn.co.za | 38.72 | 7.53 | 43.04 | 2017-10-26 06:05:27 |
Seems average-ish.
Next up, I pulled about 20GB last night. Torrents are queued, and 5 at a time are download. The first batch of downloads initiated at about 00:10, with 19 files. Downloads 1-17 completed at regular intervals, starting at 00:13 with the final one at 01:46. The final two ended at 03:14 and 05:42. Chances are the final two had a low number of seeders/leechers.
Finally, the average download speed weighted by filesize was about 1.42MB/s, i.e. sums of [filesize^2 / (endtime - starttime)] / sum of filesizes.
Altogether the performance seems satisfactory. I could probably tune my scripts to swap out low speed torrents to up performance. Also, logging max instantaneous speed would be great as well, along with upload speeds.
I doubt I'm being throttled. Those of you who suspect you have been: do you remember how much you pulled the month prior to throttling? Its possible there may be some trigger, placing you in the "throttled" category.