While it is not usually a problem with low bandwidth oscilloscopes, the relationship between the -3dB bandwidth and transition time depends on the shape of the passband so measuring only the transition time may not be sufficient.
Actually I did it properly also, with ETS and all, real -3dB point is 140MHz, cross checks with pure sine.
https://www.eevblog.com/forum/testgear/picoscope-2000/msg1153233/#msg1153233
Mentioned 200MHz with square on purpose, because Z-box owners often do not have 100MHz+ high fidelity sine sources and will do their bw testing with square. Point being that if classical frontend unhacked 100MHz will do -3dB at 200MHz with square, then 50MHz would probably do -3dB at 100MHz etc...
This does not follow. I did the bandwidth calculations based on Dave's reverse engineered schematics and posted the results on the forum (good luck finding the post though) for the DS1000Z series and they have no relationship to the maximum bandwidth. They agreed closely with the specifications.
If you want to know the bandwidth with different settings of the bandwidth limit filters, then measure them. Measure the transition times and transient response as well.
Unfortunately as you point out, most DS1000Z series users are not equipped to make useful measurements of bandwidth, transition time, or transient response. Back when this was an issue for me as well, one of my early acquisitions was a sampling oscilloscope so that I could calibrate my sources used for these measurements.
Meaning its not like you cannot measure timing related things @100MHz quite ok with any proper 50MHz scope.
Sure you can. (1) I do not use my 50 MHz oscilloscope very much except for crushing zombies but often use my 100 MHz oscilloscopes to measure timing (but not transition time) down to 1 nanosecond.
In short: this whole bw hackability thing is bit overrated in this case (especially considering Sinc trickery). Real point of hacking is IMHO more in getting all other stuff unlocked.
I am aware of the sinc trickery in these oscilloscopes but I suspect it has more to do with Rigol sacrificing accuracy for update rate.
The reason I suggest making both measurements especially with a DS1054Z hacked to 100 MHz is that if the hacked passband is not a single pole rolloff which is likely, (2) then its 3dB bandwidth will not match its transition time based on the 0.35 rule. Lots of much higher bandwidth oscilloscopes are this way but it is unusual at 100 MHz.
(1) "A 50MHz oscilloscope cannot track a 5ns rise time pulse, but it can measure a 2ns delay between two such events." - Jim Williams, Linear Technology application note 47, page 20.
(2) I base this off of Rigol's transistor selection and topology of the input amplifier shown in the reverse engineered schematics that Dave provided. The front end has an unusual arrangement of switched equalization that implies Rigol had problems meeting the 100 MHz bandwidth specification. I would expect this to result in a peaked response compromising the transient response among other things and some of the published transient response tests do show something unusual going on in the bandwidth hacked oscilloscopes.