I was flying the KOLV RNAV RW36 approach as published, approaching the iaf DOCAP from the northeast. I loaded and activated the approach in the G1000, and all seemed to be well:
Indeed, the autopilot seemed to recognize it was entering a hold and correctly started a parallel entry:
However, when the time came to make the appropriate 225-degree left turn to become established on the inbound course of the right-turn hold, something bizarre happened.
It started the left turn as expected and everything was fine until it had turned 180 degrees
But once it had gotten to exactly 180 degrees, the CDI needle immediately flipped to point back outbound and the autopilot started turning right!
I let it run for a while to see if it was just doing some kind of correction, but it just kept doing these weird S-turns and getting farther and farther away from the fix. Eventually I turned off the autopilot and guided it back to the fix myself, then manually activated the next leg in the flightplan. You can view my track log and debrief here - disregard the altitude data there, it's wrong, but not relevant to what I'm trying to show. This was in the stock G1000 Cessna 172, on a brand-spanking-fresh installation of X-Plane 11.32r1, but I've seen it in older versions as well. No additional plugins or scenery, and the only change I made was installing updated navigraph data and calibrating my joystick. I have been seeing this issue in other planes using the Laminar G1000 as well, and with other published holds-in-lieu-of-procedure-turns. I additionally experimented on trying to enter the hold from other angles to trigger different entry procedures to see if it was limited to the parallel entry. Happily, a direct entry worked beautifully, with no real problems at all. But attempts at teardrop entries (from the northwest, in this case) also went wonky, though I forgot to turn on my track log recorder for that experiment and so have no record of the exact problem.