Find your latest guilty conscience interesting... Let's try and analyse this over multiple dimensions with the goal perhaps to see where and how etc.
The table below is a rough attempt spending a few minutes only, perhaps someone can expand further based on this. For sources I only give what is available to anyone mostly free electronically and that I use frequently, or are well aware of.
Era Est #profiles Sourced% Main sources
1600-1700 Ten thousands Very High Various, main sources, not all primary, FFYP, MOOC archives, archives of Cape translations, secondary PDF documents reasonably accessible, past research/ers
1700-1800 Hundred thousands High MOOC, EggSA (GSSA) (past research/ers)
1800-1840 Hundred thousands Weak Baptisms and marriages in Cape, GSSA
1840-1950 Ten Millions Medium Death Notices from archives, baptisms and marriages on Familysearch, GSSA (becomes less over time and some families are not fully researched as (genealogical) time goes on)
1950-1970 Millions Weak Same as for 1840-1950, but much less due to not all published online
1970 - Millions Weak Even less than previous, word of mouth mostly (currently), I don't even have sources for my own grandparents or parents! :)
The above numbers and percentages are estimates, based on own experience.
I have no guilty conscience, when I touch a profile I know it is correct afterwards (for the time being and based on current information, that is).
When a profile is "done" it should ideally have 3 sources linking it altogether, I.e. and for example, it is sourced by a DN from a parent, a DN/baptism/marriage for he/herself, including name for spouse, and a baptism/DN for one or more children showing set of parents. I found that for women this is almost always easier as their marriage (and even the name of spouse) would in most cases be recorded on a DN of a parent for women. We will never get 100% for all profiles as not all DNs were filled in completely, correctly, a combination thereof, or even never filed.
There are so many life events that destroyed sources indirectly, for example 2+ marriages, children dying early and way back no DNs for them, children written out of wills...
Finally, where Afrikaans families is concerned, in earlier eras pre +-1920, the names of the parents+grandparents, and children should correspond per custom. Although there is no official source for this, it is the golden line.
Overall, there is not always sources for everything, but common sense should prevail, e.g. a scarce surname in a low populated area.
One final gripe I do have is blindly adding sources to the incorrect but similarly named profile. On a daily basis I have to address this by spending hours to untangle - buttom line is sources should not be taken on face value only.