1: You can always use MUCAPE. Typically, without an inversion, the most-unstable CAPE will be equal to the surface-based CAPE, since surface parcels will have the highest instability. MUCAPE is typically used for overnight/elevated convection just because it is in those instances where the SBCAPE can be most misleading, so to speak.
2: I'm not really sure how to answer this question, but I think the answer is no. Usually when referencing the cap "breaking", people are referring to surface parcels breaching the CIN and ascending past through LFC. Elevated instability, seen via the MUCAPE, can exist above the cap and generate convection regardless of any cap.
3 & 4: MUCAPE and MLCAPE are different ideas addressing different issues.
MUCAPE is used when there may be inversions limiting the extent of surface-based instability, but where instability may exist above this inversion. It only seeks to locate the maximum CAPE possible lifting any parcel.
MLCAPE, on the other hand, is the CAPE found when the lifted parcel has the conditions of the average lowest 100 mb of the atmosphere. This is done to obtain more realistic CAPE values, as due to low-level mixing a parcel will typical not have the conditions of the lowest (ground) level but will be representative of the lowest ~100mb or so of the atmosphere. Often, temperatures will be superadiabatic right at ground level (and therefore deceptively high compared to the T a parcel would have), and moisture pooling will often exist in the lowest data point, whereas a parcel would be much more thoroughly mixed.
Therefore, MUCAPE isn't inherently more unstable than MLCAPE.... it depends on the atmospheric situation at play. As per #3, MLCAPE can be the most unstable if the atmospheric conditions allow for it, but it isn't always the case. Check out this sounding showing elevated instability:
As you can see, the parcel ascending from 940mb has quite a bit of positive buoyancy and CAPE (and would be considered MUCAPE, as it is not surface-based and is the parcel with the most CAPE out of any in the sounding). In contrast, the MLCAPE, taken by using the average parcel conditions in the lowest ~100mb, would be T=~15C and Td=~4C, and would not result in very much MLCAPE, and certainly less so than the MUCAPE.
Keep in mind that MUCAPE will lift parcels from anywhere in the sounding to find the one that is the most unstable, while MLCAPE is essentially surface-based CAPE, except that the "surface-based" parcel has the average conditions in the lowest ~100mb as opposed to the true "surface" values.