Reproducible 2016.5 nCache missing frame issue
 

Could someone please try to reproduce this issue? I am planning to file a bug, but we know Autodesk isn't fantastic about following up.
In 2016.5:
1. Create a very simple nParticle emitter
2. Emit a few spherical particles over 100 to 200 frames
3. cache the nParticles
4. Play back carefully the nCache and watch for frames that are repeated, followed by a jump. Essentially certain frames are being read as duplicates.

If you do see this issue can you let me know:
a) Which frames you see it on?
b) Any idea what might be causing the issue?

I've had this problem across multiple computers, projects, scenarios etc. etc.

This is problem with READING the cache data, because if you open the same file with cache in 2016, it plays fine. Similarly, we've had files created (and cached) in 2016, that manifest the particle bug in 2016.5.

I don't have time to go through every project I can get my hands on that uses nParticles, but I haven't found one yet that didn't have this frame issue, given enough frames.

Thanks!

We've developed a workaround that involves changing the scale of the cache to 1.0001, which seems to resolve the issue, and should only change the simulation frame after 10000 frames, presumably.
It appears this is a rounding error? Or some kind of imprecision in the data, which is strange because the problem frames don't seem to come up evenly and are dependent on the scene/simulation.