AEC Dimensions instability

[aranoff]

Sometimes AEC Dimensions appear incorrectly (e.g. dimensions to individual windows dissappear after reopening a file) and only reappear correctly after moving the AEC dimension string and then undoing the move.
In other cases AEC Dimension strings "forget" previous edits or extend dimension strings to new points that were never chosen and again only return after moving and undoing the move.

Is this a bug or perhaps something set incorrectly?
If it is a bug, is there a simpler workaround than frequent moving and undoing?

This is a problem that has been recurring for years, though I did report it back in 2015 and support couldn't reproduce the issue in the problematic file I sent.

Please let me know if anyone else has encountered the problem, and, of course, if there is now a solution.

Thanks!

[pendean]

Does a REGENALL fix the problem when it occurs?
How frequently does it occur? Never seen it here.
Can you reproduce it on demand?

Your ACA year version is...?
OS is...?
If you turn off hardware acceleration in GRAPHICSCONFIG command does the problem go away all the time after that?

[aranoff]

No REGENALL does not fix the problem. Turning on/off Graphicsconfig has no effect.
Cannot reproduce on demand. Just tried opening the problem file I sent you from 2015 with ACA2019, but the problem does not appear.
It occurs at random times, sometimes frequently (like every few minutes; sometimes not, like every few weeks).
ACA 2019 Win10Pro
Thanks VERY much for trying to help on this!
(The only way I can reliably use AEC dimensions, especially to get correct plots, is by grabbing all the dimensions and moving them, and then Undoing the move - if not, a fully divided dimension string can appear as a single undivided dimension, with no dimensions, say, for windows and doors. Though even moving and undoing is not foolproof since the problem might possibly recur right before plotting.
Alternatively, I can explode the AEC dimensions string which keeps the correct dimensions but loses associativity, of course. Particularly frustrating is that some strings are fine while others not, so that the problem is not always readily apparent.)