Before going on to look at timelines in more detail, a cautionary note about data that tends not to be suitable for timelines.
Certain types of data are not suited to timelines, as the data does not vary over time. Common examples include:
One of the features of a unique identifier is it intentionally does not vary over time, for example each person might be assigned a unique social security number. The social security number should be modeled as a Number, not a Timeline<Number>. By contrast, a person's name may vary over time, e.g. due to marriage or change of name by deed poll, which is one of the reasons it is a poor choice as an identifier (along with lack of uniqueness);
Some data intentionally captures data which applies to a particular date only. For example, a data item for surnameAtBirth should be modeled as a String, not a Timeline<String>. A data item for incomeAtRetirement should be modeled as a Number, not a Timeline<Number>
As such, for data items of type Date, the data may be corrected in the system, but never succeeded. Thus there may be a correction history for the data item, but this correction history (i.e. the dates on which the data item was created) is rarely of interest when creating CER rules.
Typically, data types used in CER rules should model real-world circumstances, rather than corrections to the system representation. Use Timeline where those real-world circumstances can change over time; and do not use Timeline where the real-world data cannot vary over time.