Aha!

Seems I found a reason - I think groovy incorrectly handle BOM in a file header.

So, after I changed my powershell script to explicitly specify ascii/default encoding for the text output then jenkins starts to work correctly.

Thus, here is my change in powershell script:

    #[...]
    $PSDefaultParameterValues['*:Encoding'] = 'Default'
    #[...]
    function versionToFile($fn) {
      " = Saving version# to [$fn] ..."
      New-item $fn
      "dummy=123" | out-file $fn -append -Encoding ascii
      "FullVersionStr=$env:FullVersionStr" | out-file $fn -append -Encoding ascii
      "ReleaseVersionStr=$env:ReleaseVersionStr" | out-file $fn -append -Encoding ascii
      "BuildStarted=$env:BuildStarted" | out-file $fn -append -Encoding ascii
    }

Now the same jenkins code generates following output:

    * loading ini: C:\XBuild.main\build\log\ARC\version.txt ...
    [Pipeline] readFile
    [Pipeline] echo
     = ver: 3.3.0.160; props:{dummy=123, FullVersionStr=3.3.0.160, BuildStarted=2025-04-06,16:22:12.944, ReleaseVersionStr=2020.1.1.1}
    [Pipeline] echo
     = ver.a: 3.3.0.160; 
    [Pipeline] echo
     == (false) [dummy]: 123
    [Pipeline] echo
     = set vn: 3.3.0.160/3.3.0.160; 
    [Pipeline] echo
     == (true) [FullVersionStr]: 3.3.0.160
    [Pipeline] echo
     == (false) [BuildStarted]: 2025-04-06,16:22:12.944
    [Pipeline] echo
     == (false) [ReleaseVersionStr]: 2020.1.1.1
    [Pipeline] echo
     = ver.b: 3.3.0.160; 

So, now all methods of properties reading works fine!

That is very interesting.

As you can see - there are only latin letters are used and even with classic latin letters string values are compared incorrectly.

For sure - there is bug with handling Unicode files but the question is - if this bug on JDK side or in groovy?...