Two categories of time:
There's the time that we normally correlate to the time we see on our wristwatches, phones, computers, wall clocks, etc. Lets call this "wall time".
Then there's the time that correlates to actual time, which accurately and uniquely represents any point in time, in a fashion that allows us to do plain arithmetic. Lets call this "real time"
In Cosmos, the "wall time" is natively represented based on modified julian day, with a resolution of microseconds. The benefit of this is that all calendar dates for any recorded event till now (and thousands of years into the future) can be represented. And even the time of day for all those dates, are representable down to a millionth of a second, which seems like a practical resolution for general purposes.
These "wall times" will generally be named in terms of calendar years, months and days, hours, minutes and seconds. Also the concept of weeks is rather common. Both in terms of UTC (Universal Coordinated Time) and geographically determined local times.
The primary issue with "wall time" is that we keep that wall time in sync with earths rotation, which requires the insertion of leap seconds. These leap seconds are generally entirely ignored when we deal with calendar dates and times, because we as humans will barely notice the difference.
But in precision systems where timing is paramount, it could easily be a disaster if time jumps a second forwards or backwards. Fortunately, the calendar date and time of day, is not of much significance in such systems, as long as the relative time among the systems can be determined. So if some event is recorded one second before a leap second occurs, and one second after a leap second ocurred, three whole seconds of time will have passed, even though in terms of calendar date and time, there's only a difference of two seconds.
In Cosmos this kind of timing information is natively represented based on International Atomic Time. The benefit of this, is that any event that could have possibly been recorded at a uniquely identified date and time with sub-second accuracy, will be representable. These events are representable with a nanosecond resolution (billionth of a second) for more than 500 years into the future.
January 1st, 1958 till 2542
It is possible to convert any atomic time to wall time, although with loss of information. It is therefore not necessarily possible to convert the wall time back to atomic time. This conversion also requires the presence of leap second tables, to limit the loss of precision to less than a millionth of a second.