Ok. I got it working!
I am accustomed to running everything from a profile in Ekos. So starting it from terminal was a challenge for me, and the folder /var/run/indi does not exist on my system. How can I add it to the list for focusers in the Ekos Profile Editor?
Onto my tests.
The motor drives very well. But once it reaches a position there is a hum.
Parking the motor via driver disconnect is fine, and power off properly, no hum.
I had no issues with speed set to maximum. Could even go higher.
100 Backlash steps is not enough for me. Will need at least over 200. Please increase to a generous 500 step limit.
It may be interfering with USB devices, such as WiFi dongle. Not necessarily the driver's fault. Will have to test further with more equipment attached.
I noticed a bunch of messages in the terminal that started up the indiserver that coincide with my WiFi dongle cutting out momentarily. They read

Thank you

Just copy indi_amh_focuser.xml file to /usr/share/indi - if it's not there. If it's there you can select a driver from profile editor in Ekos.

Does a stepper motor get warm after a while in the stop position? If so, it means it uses break/cost mode to hold the motor still in the last position. I will disable if it is the case.

500 is a lot for a backlash. Could you check how many steps you need for 1 revolution of the motor?
I use DOUBLE step mode in the code (see lines 456, 461, 469, 474) - could you change these to SINGLE, recompile and check number of steps for 1 revolution again?
Use absolute position only for these tests so just set the required position to a number and let it travel to this position.

I will not be able to do anything with this. It probably comes from wiringPi library OR it might power drop issue if you use raspberry pi to power your motor.

It isn't working hard enough to get warm. But I can monitor the current it is using. The current remains elevated while in stop position, drawing an additional 0.2A at 12V. It definitely resists being turned in that state. I would agree with 'break/cost mode'

I opened it up to get a better look at the main gear shaft. My excessive backlash was mostly due to my 3D printed gear slacking up. I shimmed it with something and now backlash is only 20 steps. I will probably add a proper setscrew next time I can reprint it.

1 full rotation of the main shaft is 10212 steps through the planetary gear reduction of ~51:1

It appears this was an issue while connecting to the Indi Server via KStars Lite. (A step I took while trying to work out how to connect in KStars, to verify it was running.) It wouldn’t be due to power drop simply because the Motor Hat has its own separate power supply. I will monitor this as I go.

As far as I can see from your previous post in this thread you only need 1,75 focus knob turn to travel from minimum to maximum. To make sure you get required precision you need to calculate actual travel of your focuser (in millimetres) per step. This value should be within critical focus zone value, which you can calculate here . Depending on your results I need to use proper stepping mode in the code (single, double, interleave, microstep).

There's no dedicated function for disabling break/costing mode in the library but I will come up with some solution.

1 Full main drive gear 10212 steps. Increased by gear on focus knob by 20:25. Factor of 0.8
8170 Steps for 1 full knob rotation. [10212*0.8]
14297 steps for full travel [8170*1.75]
Total Focuser Travel 40mm
0.0027mm/step [40mm/14297 steps]
2.7 microns per step
Critical Focus Zone 62 microns. Or 0.62mm

This should work well

I was looking around for some examples of code for this and found one with a variable for Break. This is not my expertise, but maybe it's an example for what you are looking for.
github.com/adafruit/Adafruit-Motor-HAT-P...t_MotorHAT_Motors.py

Good point!. In the python library BRAKE and RELEASE modes are explicitly defined and are available. I just noted they are also declared in C++ library . I will try using RELEASE in the driver's code and let you know when it's ready for tests (probably over weekend).

After closer look I found out that RELEASE function is not implemented in C++ library. Therefore release is achieved by resetting all PWMs after stepper motor finishes its travel to destination. You can checkout the latest code from github for testing.

Today I mounted everything together and took the motor for a proper spin. I can confirm 14,000 steps is all I need for full travel.
The process takes about 225 seconds. It's rather slow, but I can live with it as my prime focus is mostly in. But if it is safe to increase speed, that would be nice.
I also found more backlash in my complete drive train. 100 doesn't appear to be enough after all, 300 or 400 should do the trick.
Final thought, as a driver for DIY focusers, other users may desire the ability to select within options the drive mode (Single, Double, Interleave Microstep) that best suits their choice of hardware. For instance if somebody chooses a stepper with 5:1 or even no gear reduction; they would probably benefit from Microsteps.

Excellent. I'll have a look right away.

I'm glad everything works for you.
The speed value cannot be higher than 255 and now the driver lets you set maximum 250. So it's rather the problem of high gear reduction ratio for the motor than slow speed. I will increase maximum backlash right away. Checkout the sources in 5 minutes.
Let me add selection of various modes in a future release.

It looks like resetting the PWM did the trick. Nice work!