Sudden motor jump after machine enable. Delta ASDA e3 drives and EtherCAT

So I've got my machine to the point of wiring up and testing motors. Made adapter plates and fitted everything off temporarily.

I'm running into an issue when I hit machine enable in LCNC.
One or more motors will jump, move that is, violently for a split second.
Only for a split second and only once, a few seconds after I hit enable.
Sometimes it'll cause LCNC to error out (servo amp error?) and other times not.
The machine continues to work as if nothing's happened though, if it throws an error i just hit enable again and it's good to go.
No jump after the first startup. I can turn the machine enable button on and off as many times as I like and it's fine after the first time.
Sometimes it doesn't do it at all!
I've checked all the encoder cables and motor cables.
I've swapped all the cat5 cables connecting the drives.
I've auto tuned all the drives in ASDA soft. Any jogging or tuning functions from the drive or the drive software work flawlessly, no jump.

I'm at a bit of a loss. I assume LCNC is sending something to the drive and maybe it's position dependent? I've messed around with how I close out of LCNC looking for patterns but I can't really narrow it down.

I've uploaded a video of the Z axis jumping. It's pretty violent, like max speed and acceleration for 1/10 of a second.

Youtube Video

Hal, ini and ethercat xml are attached. I worked from someone else's files, hence the EL8 name.

Ok so I've found what's happening but not why. Video below shows it in detail.
A few seconds after Lcnc loads the, I guess, current possition of the machine, the possition it starts up in lets say, is changing. I can see it jump from one position to the new one. Sometimes it's just the Z that jumps, sometimes it's the z and the y.
I hadn't noticed it before as I was so busy watching the motors. If I enable the machine before the current position in lcnc updates it tries to move to that new position. Again, easier to show it than describe it. It's pretty clear in the video, linked here

Could this be something to do with absolute encoders? The original machine that I'm using the files from had absolute encoders. Is there something saving the machine position on shutdown or something like that?


 

Looks like it's going to be the Home_absolute_encoder=2
I should have said that my motors don't have absolute encoders.

Ok, changing the Home_Absolute_encoder=2 to =0, no change. #'d it out altogether, still no change.
What seems to be happening or at least what I can see is that when lcnc starts up it's loading the stored g54 values from linuxcnc.var as the current machine possition but then it's zeroing out the dro a few seconds later. If the machine is enabled it's actually trying to make that move to 0 rather than just the dro zeroing out.
I've no idea at this point. Maybe an actual homing routine will sort it out. At least for now I know that if I wait a few seconds before enabling the machine I won't snap any motor shafts.

It could be interesting to use halscope and record pos-cmd and pos-fb around the Machine on event.
I think your analysis is right. I have a similar problem. I zero the encoder's values before starting linuxcnc.
My issue is related to stepper motors when energized, so not necessarily same for you.