This is a tricky question (besides harmonics there is interference issues). Simple answer I have not, but then again I've not had the opportunity to fix an issue.
Several things should go into the decission before buying a reactor, for just placing in reactors in several locations on the same source can cause problems through harmonic interactions and you end up with worse problems than what you started with.
So, how much of the locations load is VFD's? How big and how many are there? What is the patern for operation, so maybe an interactive unit (assumeing there is such a beast) is needed.
It comes down to needing to get it engineered by an engineer with experience in the area. I am not one with the experience in this area and my limited knowledge almost pre-dates VFD's.
TC-ER cables or XHHW-2 conductors in metal conduit may be all that is needed thus sheilding other things from the interference.
It's a big coil of wire. Half a transformer. What can it do?
All it can do smooth out 'bumps' in a less-than-perfect DC supply, and alter the impedance of the circuit.
So, any effect would appear as better speed control and cooler motor windings.
With that in mind, I have no doubt there are plenty of folks who routinely bypass them, and claim things work just fine. For a similar reason, I often find braking resistors disconnected. Who needs them?
Well .... let me put it this way: Every time we get a factory rep out, he makes problems disappear, usually by simply restoring everything to its' original configuration. That's followed by a year or two of 'field improvements,' a complete failure, return of the factory rep, and the cycle repeats.
Another common thread to problems is the equipment being run way beyond its' intent. 12-ton cranes are not really meant to lift 12-tonnes all day long, but the company 'saves' by not buying the 15-ton crane. Small wonder that two years later the crane (or whatever the machine) is thoroughly hated five years later by the maintenance crew.
The need for reactors cannot honestly be answered by the manufacturer at the time of purchase. Except for extremely cheaply engineered VFDs, manufacturer recommendations for reactors are usually based on the philosophy of 'better safe than sorry'.
The need and benefit of reactors can only be determined determined by the installed system. Installing reactors into existing circuits is harder than putting them in during the initial installation.