For example, consider you are flying the 8KCAB Super Decathlon (a fairly common acrobatic plane) that had G1000 (using the flagship Garmin glass cockpit as an example, answers for other types of GA glass are welcome) avionics fitted to it, and you were about to go off and practice your favorite aerobatic routine in the local aerobatics box. Would you have to do anything equivalent to "caging" in order to keep the attitude indication functions of the G1000 from misbehaving during aerobatic maneuvers, or is that not an issue with the types of "glass cockpit" instrumentation used in the GA world?
Caging was required when Gyros were mechanical masses actually spinning at many rpms in their cases. That mass resisted rotational motion and produced a resistance force that was measured and translated into useful info like pitch and roll. That spinning mass carried much momentum in the form of its mass times its rotational velocity. And too much sudden motion and the resistive force/momentum could cause mechanical stress to the unit and/or cause the instrument to tumble. So it became helpful to cage a gyro before acrobatics to keep the thing from busting or wearing out prematurely.
In modern glass cockpits like the G1000, there are no mechanical gyros with high mass parts (excepting a standby instrument independent of the glass system, which may need caging if it's a mech type). The gyros that are used for a G1000 system are called MEMS (micro electro mechanical system) gyros which do not have a spinning mass. They have tiny metal discs that merely vibrate. When they are rotated (by pitching or rolling), the vibration dynamics change a little bit and a capacitive reactance apparatus measures this change which is directly proportional to the rate of rotation.
The MEMS gyros can take aerobatics without enduring any mechanical stress what so ever so there is no need to cage them.
The system does however, loose accuracy. The rotation rates offered by the MEMS gyros are prone to micro errors and these errors add up over time. The more rotations, the sooner the sum of all errors becomes significant. There is a software algorithm that re-aligns the system periodically and basically erases the errors (it kind of reminds the system which way is up). The algorithm uses accelerometers (like the one in your phone) to sense stable, un-accelerated motion, and at that time, the accelerators can sense gravity which points straight down. When this occurs, pitch and roll angles are keyed off the accelerometers and the rotation rates are reset to zero and the system is ready to start keying off the gyros again and adding up errors. In normal flight like touch and goes or cross country flight, there are lots of opportunity to reset the errors. In aerobatic flight there are fewer opportunities but once the aircraft eventually returns to level flight for some period of time, it will reset itself. Whether or not the error can be significant enough to be noticed by the pilot, I don't know... probably depend on many factors, but I'd probably be intimate with a sic sack by then.
I fly a Super Decathlon with an Aspen E1000. While an excellent PFD, I certainly would not trust it in aerobatic flight. There's no cage function, and no way that I'm aware of to keep the attitude indicator from misbehaving, which I've seen it do firsthand.
I also fly the Avidyne R9 system, which also does not have a caging function.
Furthermore, I downloaded the Garmin G1000 manual, which says nothing about a way to cage the GRS 77.
This lack of commonly available GA glass-cockpit aerobatic-capable instrumentation probably contributes to why one does not do aerobatics in IMC.
Whether caging is required depends on the system. Some glass cockpits use conventional gyros which can tumble and mis-align, while others use accelerometers, solid state gyros, or laser ring gyros which have no mechanical parts and do not require caging. I believe the G1000 uses solid state gyros.
That does not means that you won't get erroneous readings from a non-mechanical glass cockpit during aerobatics, processing power and software design are other factors that come into play. It does mean that you should not have alignment issues once you are done.