PSG ascent guidance coast staleness

[Krakon01]

MechJeb Version

2.16.0.0-1703

Description

Using PSG ascent guidance, when coast stage is enabled, a few seconds before the coast stage fires staleness starts going up and never converges again.

[lamont-granquist]

That isn't really useful. I've been doing tests on Kerbin debugging the "coast during" burn-coast-burn behavior and it works reasonably fine (#2272 #2284 #2285). If you want to submit a bug report, I need the craft file, and at least a shot of all the settings in order to replicate it. Bonus for any sort of video of the launch.

[Krakon01]

I have recorded a video of a launch, right before coast stage lights up staleness starts going up. I'm playing with RP-1 and Principia. Here's the link to the craft file.

[lamont-granquist]

Okay, so the important point is that you're using Coast Before checked with the coast stage (2) equal to the burning insertion stage (2). The staleness increasing 10 seconds before the end of the coast is normal, but having it never settle again is not. The fact that the coast stage is still suspending the optimizer after the coast is over, is a bug, but the fact that the coast stage matches is important in figuring out how to track it down.

[anmpereira]

Found this bug today too. Coast was 192s and stopped converging from 10s before burning insertion to the end, for a total of 3m30s.
I tried changing the following:

1. Coast max 180s, still bug
2. Cost max 90s, still bug
3. From ("coast stage = 3", "Coast before") to ("coast stage = 4", "Coast after"), converged

Settings at pad:

Craft file (txt because I couldn't upload otherwise)
MoonOrbiter.txt