Version of Forge in Use: 1.10.2-12.18.1.2076 also 1.10.2-12.18.1.2065. No other Forge versions were tested.
Only fresh installations of Forge were used and no mods were installed.
Tested with fresh installations of Vanilla 1.10.2 and snapshot 16w33a. I have been unable to reproduce this problem with those, so I assumed it was a Forge issue and am reporting it here.
TL;DR
It appears that comparators can treat double chests as 2 separate chests depending on the circumstances. The top 3 rows are considered one chest while the bottom 3 rows are considered a second. The comparators only seem to read their "side" when the chest is emptied with a hopper. A block update is required to turn the "stuck" comparator off. If comparators are placed on both sides of a double chest they both properly detect when an item is placed anywhere into the it, but only one comparator properly detects when the chest is emptied with a hopper.
Full Description and Screenshot
Here is a very easy method of reproducing this problem.
If you place items into the chest, both comparators will turn on and output an expected, proper signal. When a hopper under the chest removes all of the items, only the comparator associated with that "side" of the chest will update and turn off properly. The other comparator will remain on until the player manually causes it to update by either opening the original chest, breaking the comparator, etc.
You can verify this behavior by alternating which rows of the double chest you use. If you place the items manually into only the top 3 rows, you will find that the comparators produce different results than if you had placed them in one of the bottom 3 rows.
Though my screenshot does not show it, it does not seem to matter which side of the double chest the hopper drains from, only from which "side" of the chest the last item was removed. I tested all 6 adjacent sides of the chest and the problem occurred the same way each time.
The orientation of the chest being measured by the comparators also does not appear to matter. I built chests rotated in all directions and facing all directions and the results are consistent. It does appear that the North/West half is considered to provide the top 3 rows, and the South/East half provides the bottom 3 rows.
I hope I have provided enough useful information.
Thank you very much for reading.
