I read some info about this caps and the interesting for me is that Tektronix exchange t11 (8.2uf/75v)caps with BMIcaps aluminium (500d series 10uf/100v) and after this with Nichicon aluminium (22uf/100v) TLB2A220TCAANA. This is in the power rail of +/-50v. In 5V rail, they start to put inside tantalum (100uf/20v - tdc107m020wld) and in the end of production put Panasonic electrolytic ECEA1EV221 (220uf/25v).
I have been compiling notes on the 7000 series switching power supplies to track how they evolved over time because I have several which need to be refurbished but that has not included the bulk decoupling capacitors on the interface board.
Your observations are consistent with the ESR and frequency performance differences between tantalum and aluminum electrolytic capacitors. I suspect Tektronix found that the original 10uF replacements had too high of an ESR at high frequencies and switched to 22uF to solve that which is inline with my rule about replacing tantalum capacitors with aluminum electrolytic capacitors. Replacing the 100uF/20V tantalum with a 220uF/25V aluminum electrolytic is consistent with that as well.
So it was the ESR at high frequencies which matters as usual for a bulk decoupling capacitor.
I do not know why Tektronix used wet instead of solid axial tantalum capacitors for the 50 volt supplies early on. They used solid axial tantalum capacitors elsewhere for the same application before replacing them with aluminum electrolytic capacitors.
Now put low esr aluminium caps from Panasonic and Fujicon (this I can find easy here) and think that work well but maybe need to check more when put more plugins and to check how they will work on big frequency of the input signals. Maybe when have a time will look for ripples in this power rail when the scope work with more plugins installed.
That is what I would do. You could even quadruple the capacitance from the original tantalum value just to be sure. I would use the Nichicon PW series (105C, long life, inexpensive, miniature, low ESR, high frequency) but there are plenty of good options and it would be difficult to go wrong.
Something else for this scope. I have no calibration plugin for this series. Can be possible to calibrate this scope with cg5001 (I have it on my lab, but for now always use it with 4xx, 22xx and 24xx series scopes). In the manual I read that need to do this with calibration plugin.
Except for transient response, the calibration can be done with a vertical amplifier by connecting test points for an AC/DC multimeter to the differential signals at the interface. I use a 7A16A (1) but any of the single trace high input impedance vertical amplifiers are suitable. The dual trace amplifiers will work also but access is more difficult.
Transient response calibration is more difficult and would require a 7A19 or 7A29 and an external adapter to sample the differential signal between the plug-in and interface using the procedure Tektronix used for calibrating the 7104. (2) I just settled for verifying the transient response at the BNC with my fastest vertical amplifier plug-in which is a 7A24 and fastest reference level pulse generator; it looked good enough to me so I did not mess with it.
(1) The 20 MHz bandwidth limit of the 7A16A helps a lot by producing a low noise trace so the 7A15A with its 10 MHz bandwidth limit may be even better.
(2) Tektronix made a special plug-in extender for 7104 calibration with coaxial connections for the differential vertical and trigger signals that could be disconnected to be swapped or sampled.