I wouldn't rely on intervals for precise timing. They are pretty accurate on their own, but when coming down through the library and program, the accuracy will drop. Program event times vary (especially if used in a GUI scenario, where there's generally 10ms of overhead), so you're best off using timing from the OS.
I've seen event times vary +/-5ms to +/-10ms on random data points in C# GUI applications. But in general the average time between events will be what you specify as a DataInterval.
But to answer your question it is the TMP1101 that handles timing, and if running 4 thermocouples at minimum DataInterval, you could see the intervals go up due to high load.