there out in the field running for over a year so they work you could say!
The PIDE Instruction, by default, uses the "Periodic Timing Mode", and it is recommended that it is put into a Periodic Task. For absolute "best" results, that periodic task should have the highest priority (lowest number).
It
can be used in a continuous task, triggered by a timer (Oversample mode), or it can be executed every scan and the instruction will use the time-stamp from the process input if available (Real-Time Sampling mode).
As you say they are in periodic mode (default), and they are most likely running OK because they are triggered with a timer.
Typically there's "not much" wrong with that setup, but only in loops that don't need fast response or super-accurate control. The natural fluctuation of scan-rate will affect when a timer .DN bit is seen (in real-time), so the PIDE will not be scanning at a fixed, deterministic, rate.
My guess is - a wild guess - the SIs didn't have a full understanding of how best to use the instructions, and I'll apologise to them if they can demonstrate superior performance the way they have it. A heavy comms load, for example, can affect the continuous task scan-rate dramatically
As Alaric points out "best" performance is achieved if the instruction absolutely knows the exact time-interval from when it what last scanned, because PID uses Integral and Derivative terms that are time-based. Why settle for "second-best" ?
An analogy - would you measure anything with a tape measure that someone had put 1 metre marks on it by eye ?