Looking at the CPX200 (10A) (schematic from link above ) and a CPX400a (20A) schematic and they are radically different.
The CPX 200 a dual switched flyback (No separate post regulator).
the cpx 400a is a dual switched forward converter + magamp for pre-regulation followed by linear output stage.
I also use a TTI supply the older Pl quad series with 'hard' switch disable and remote sensing, pure linear no pre-regulation. Been a very reliable supply so far.
A couple of possibilities to consider.
These supplies (and many others) use resistors (PTC in the cpx200) to guard against remote sense accidents. But as this previous post suggests (IMHO) I think resistors are inferior to diodes as guards as they offer no protection to loads in the 2 scenario's noted there .
https://www.eevblog.com/forum/projects/korad-ka3003d-redesign-and-upgrade/msg1017611/#msg1017611 A 'guard diode' would always protect the load since it clamps the main output wire voltage to the remote sense wire voltage with a max difference of Vf ( i.e it protects the Load from excessive voltage in preference to protecting the sense wire from excessive current ).
The two scenario's :-
1 / If main wire has become disconnected (or similarly sense wires are connected first before main wires) then the main output terminal voltage 'Vmain' can rise up to to Vmax, this can happen because any current flowing through the guard resistors to the load causes a voltage drop 'Vrguard', so now Vmain = Vsense + Vrguard, any large output caps also get charged to Vmain. These will then get discharged into the load when the main wires are finally connected.
This may be occurring when your 'hot-plugging' which is connecting both the Vsense wires to the load a few mS before the main wire makes contact to the load. In those few mS Cout could charge to near Vmax.
The time constant of the CPX200's output caps and three parrallel output resistors is 3000u*400 = 1.2s thus for them discharge from 40V to ~ 2V = 3.6s , is this the 3 Seconds over voltage your referring to ?.
Another scenario that 'guard resistors' offer no protection against :-
2/ Mixing up polarity of sense wires . Again Vout will go up to Vmax for same reasons as 1 above and will remain there (if supply has an OVP trip that senses over voltage at the main wire (but not at the Vsense node) then this should activate and shut supply down) .
Is it possible to accidentally re-enable a bench supply .?
Well yes if it's one that disables it's output simply by setting Vset to 0V (a 'soft' disable) then consider it still in a normal 'on' state, it's just that it's output is now regulated to 0V . So as with any 'on' supply if we confuse it by applying a voltage to the Vsense terminal it will react. So a brief negative voltage applied to it's Vsense terminals (as would happen in a reverse polarity Vsense connection or Vsense connected to a negatively charged cap ) would cause the output voltage to rise sharply an attempt to maintain it's Vsense terminals at 0V.
But if that schematic linked in this thread is the correct one for your CXP200D I can't see how the output could be accidentally re-enabled by the method above since it looks like the output disable operates by shutting down the PWM oscillator that drives the flyback (shutdown by mcu).
A quick test :- with supply output disabled and in remote sense mode , charge a largish test cap (~-1V) and connect it to ONLY the Vsense leads (such that +Vsense and -Vsense will momentarily be reverse polarity ) whilst monitoring the main Vout terminals with scope.).
regards