Again, thank you for the swift reply.

Your answer concerns me though, because if I'm correct, the general reliability of SF-TDA for solution-phase reactions becomes questionable.

Could you maybe recommend a route to communicate about this topic and find solutions? I do know that a lot of ORCA discussions are ongoing in Keinsci forum, but I'm not confident that an English thread (or a poorly translated Chinese thread) will be taken seriously there...

I have tested an organic molecule (neutral, singlet) in ORCA's spin-flip (SF-TDA) calculation.

Below is the general input command block that I used:
! BHANDHLYP def2-TZVP tightSCF CPCM(DMSO)
%cpcm     xfeps          0.5     end
%TDDFT
  SF TRUE
  NROOTS  10
END
* xyz 0 3
...

As a result, the spin-flip ground-state was the singlet (as expected), and the spin-flip first excited-state shows S^2 of roughly 1.1, which seems to indicate the singlet diradical species.

My question is the following: When I do SF-TDA, the initial reference state is triplet, not singlet. Then, does the "slow" portion of solvation (that is fixed for non-equilibrium solvation calculation of spin-flip states) correspond to the triplet? If this is true, wouldn't this be unphysical, because in reality the slow portion of solvation should be that of the singlet ground-state?