How to fix "Queue is not empty after probing"

what debian flavor?

bookworm

github.com/LinuxCNC/linuxcnc/actions/runs/25025393650?pr=3973

look here at the bottom of the page

I guess I'm not very tech-savvy. I couldn't find a link to the Bookworm DEB package at the bottom of the page (and I need to figure out whether it's the real-time kernel version or the userspace version).

github.com/LinuxCNC/linuxcnc/actions/run...artifacts/6673490803
is uspace preempt-rt

Thanks, I downloaded and installed it.
The latest changes to my configuration were as follows: I have a machine with a rack-and-pinion system. It has a good control board and cable. The XXYYZ configuration, Matrixkins—I backported it myself from the source code, adapting it for 5-motor control. The configuration has been tested and works.

To ensure the minimum Z-axis step size was acceptable for scanning, I switched the step divider from 8000 to 25000. For my speeds, I had to reduce the value
[EMCMOT]
#edited 2026-03-21 due to jitter tuning not exceeding 2800 (previously 8000) - base_period reduced from 20000
EMCMOT = motmod
COMM_TIMEOUT = 1.0
BASE_PERIOD = 10000
SERVO_PERIOD = 1000000

What did this lead to? Version 2.8.4 with these same parameters works like a charm. The version you sent displays a message at startup - -
Unexpected realtime delay on task 0 with period 10000
This Message will only display once per session....


then goes through homing

+ and when attempting to move the Y-axis, the Y-motor emergency protection trips (torque overload). Obviously, something is wrong with my Matrixkins. Apparently, my participation in testing is causing more problems than it’s helping.

what are you using the base-thread for?
10000 seems incredibly small.

if you are using custom kinematics, you might have to build again from source, you may not be able to use my debian packages.

It requires you to be logged into github to get the debs I think.

The default setting of 20,000 corresponds to a maximum pulse frequency of 50 kHz. An LPT system with a high-quality cable and control board can handle much higher frequencies. 10,000 corresponds to only 100 kHz. I managed to stabilize the latency test at a level no higher than 2500.
My stepper motors run at speeds of up to 20,000 mm/min.

Ok, I'm not sure I can help you with the latency and the Y errors way out of scope of the probing issue, you had another 2.9 that was working ok, where you could patch my changes?