%Tt was not good, because it also included the idle time from previous
request when the connection was reused, causing times recorded in log
files to be much longer the actual request duration.

%Ta - which is described as total active time for the HTTP request,
between the moment the proxy received the first byte of the request
header and the emission of the last byte of the response body - is much
more suitable for our use case of measuring requests durations and is
similar to apache's %D