Acquisition time of Sentinel-2

Dear colleagues,
can somebody please give this piece of information concerning the (sensing) time of S-2?
In EO-Browser each scene has been given a time of acquisition. Is this the start, centre or end of the scene?
I found that in the Copernicus HUB S-2 headers there is just one time indication: Start-Sensing time and it is the same for all scenes in this bit of orbit acquisitions.
E.g.
Start sensing time (in the HUB) 14.47.31 (hours.min.sec.)
EO-browser 1st scene says 14.47.44 (14 seconds later)
2nd scene 14.47.58 (14 seconds later)
etc.
This would mean EO-Browser provides always the end sensing time of each scene!
Is this correct?
I need this information because I need the exact sensing difference to Landsat-8 of the same day which in L-8 is always giving as the centre-time of the scene.
Many thanks in advance

The information we show in EO Browser is extracted from metadata.xml file, attribute SENSING_TIME, see example here:
https://roda.sentinel-hub.com/sentinel-s2-l1c/tiles/1/K/AT/2018/7/20/0/metadata.xml

In the Sentinel-2 Products Specification Document (PSD) they describe this attribute with:

TILE Start Time. This value is currently set to the Datastrip Start Time (cf. datastrip definition in section 2.3 )
Note: set as Type date_time:AN_UTC_DATE_TIME

These information are unfortunately not as clear as one would have expected. You might also notice that Copernicus Sci Hub is displaying a bit different times (a couple of seconds off).

I suggest you contact Copernicus EO Support (eosupport@copernicus.esa.int) for more info. If you get something deterministic, we would appreciate if you share it here.