tl;dr I think we should get rid of both of these histograms, and optionally replace them with something that can actually be interpreted.

The UMA histograms Sync.SyncEverything and Sync.CustomSync supposedly record whether users have the "sync everything" flag enabled, and which data types they chose. However, the recording logic (especially for the latter) is so weird that they're IMO not very useful, and probably more harmful than helpful. In particular:

Sync.SyncEverything is recorded every time the value changes, plus during the first setup (where the previous value isn't really defined). That's at least straightforward, but not all that useful:
- For existing users that don't change their settings, it's never recorded. So it doesn't actually tell us which fraction of users have selected custom types.
- It's recorded immediately when the checkbox is toggled, so *before* the change is actually applied. If a user can't make up their mind and toggles it back and forth 10 times, we'll record 10 samples.
- It's buggy: During initial Sync setup (even without going through the advanced settings flow), it's recorded twice, once with "false" and once with "true", because PSS::OnUserChoseDatatypes gets called twice for some reason.

Sync.CustomSync similarly tries to record only when something changes, so shares all the above problems. Additionally, it only records when a data type is turned *on* - we don't record anything when it's turned *off*.
Exercise: Go to chrome://settings/syncSetup, flip one of the data types on and off a few times, then go to chrome://histograms and try to make any sense of the recorded values. Try both during initial setup and later re-configuration and pay attention to the difference :)