Running over the data again, here is an update on the Run14 swapped towers I discovered.

Short Summary:

 The first swap I found are swapped for the entire run. These are:

 3445, 3446, 3447, 3448  <-------> 3452, 3454, 3455, 3456
 3425, 3426, 3427, 3428  <-------> 3433, 3434, 3435, 3436

They are separated in eta at about ~0.4

The second swap I found are swapped for only part of the run. After rundday 106, during access these towers are swapped

2781, 2782, 2783, 2784  <--------> 2821, 2822, 2823, 2824
2801, 2802, 2803, 2804  <--------> 2841, 2842, 2843, 2844

They are separated in phi ~ 0.1
To figure out the runday, I just looked through the Run14 shift log until I found any information on the suspected towers, which for the second swap belong in crate 0x04.

Here below are the plots I used to determine which/when the towers were swapped for the second set.  Used HT2 triggered events. Plotted relative location of tower energy > 3.0 GeV and track pT > 6.0 GeV (Before hadronic correction)

Before Run 15106094 (Time of swap for second set):

--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
 
Below is after Run 15106094 (Time of swap for second set)

------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Now below are the same histograms except for the first set. The towers are separated by about ~0.4 in eta. This swap exists for all of run year 14.

 And here below is tower 3453 which serves as an example of a non-swapped tower as seen by the analyzer. This tower is known to have been swapped with 3452, and was corrected for. This explains why 3452 is shown to have been swapped and not 3453, as one would expect from the pattern.