I have checked trigger T0 offset for Run XIX  Collider samples:

       Sample                               T0            dT         ddT       freq

    {"2019/RF/AuAu200"      , 2.3904,-0.0482, 0.0201, 9383180}, 
    {"2019/RF/19GeV"        , 2.3675,-0.0023, 0.0106, 9341081},  
    {"2019/RF/14p5GeV"      , 2.3904, 0.0070, 0.0122, 9307123}, 
    {"2019/RF/9p2GeV"       , 2.3904, 0.0141, 0.0152, 9188688},
    {"2019/RF/7p7GeV"       , 2.3904, 0.0362, 0.0144, 9104517}  

where T0 trigger time taken from DataBase, dT +/- ddT is the measured difference between z of primary vertices reconstructed using West or East parts of TPC only, respectively, expressed in μs (microseconds).

I do see that triggerT0 (μs) = T0 + dT = (0.1 +/- 0.5) μs  +  (21.2 +/- 5.4) time buckets (with kind of strange point at 14p5 GeV) i.e. roughly speaking our trigger decision is coming in 21 time buckets from the collision.