We have double checked the behavior of our 7033 NIM bin.
The 7033 is specified for 2.5A at +/-24V.  (== 9.6 ohm load resistor)
The current limit (fold back type) is set to 3A

We made tests: (at the +24V line):
R load   I load    U out     remark
6.8ohm    3,54A   24.00V     OK
5.8ohm    2,54A   14.74V     fold back limit
5.1ohm    2,06A   10.53V     fold back limit

That means: The crate works also with 3.5 A although it is specified at 2.5A
If the overload (lower load resistance) is removed, the voltage restores to the nominal value!

I have another suspicion: The 7033 (and all odd numbered bins) have an overvoltage protection. Say during switching on the crate there is a current flowing from the +24V or +12V back into the +6V line due to a special circuit in your load. In this case, the overvoltage circuit would trigger and short the +6V line. If the +6V line fails, there is no supply current to the regulation circuit of the +24V and the +24V ( and also the +12V) would fail as a consequence.
After switching of the crate, the overvoltage circuit resets and everything is working again.

Possible test:
Populate the 7033 crate with several NIM modules drawing current of the +/-6V and +/-12V lines. If you add the special module, that is causing this strange behavior, may be it is working now, because the overvoltage circuit does not trigger in this case. The other modules use this backward current and there is no overvoltage.

A crate without overvoltage protection or with high leakage power consumption would probably show a normal behavior in this special case.