question about offset

Hi tommylight,
I am sorry, I am not sure I understand.
you are saying that one of my axis is set to travel half? how come? like you see on the picture three sides were cut correctly - that means to me both axises are moving correctly... the settings are applied all the time, right? so, if machine cut correctly side #1, side #2 and side #4 how it can be incorrect setting in ini file - please explain that to me.
corners were not intended to be 45 degrees, nor I care about it. I made a model in Rhinoceros. than I feed this model to program I have written to produce ngc file. this program reads from template, where all preambles for nc code are set. and it writes all these lines to final gcode file. then it reads model's lines and generate accordingly either G1 and coordinates or G2.
I found recently in linuxcnc manual that G2 may not be as precise as it might needed, so, I am planning to re-write my program to use G3 instead of G2. but for this mystery I am trying to solve now it is not that important... I think.

yes, andypugh - that is the field.

I was watching the wire all the time. and to compensate for this wire bending (so far manually), I was stopping gcode execution at the end of each line, so, the wire could have a chance to catch up. I do not have (so far) any mechanism in place to check for wire straightness so, I have to trust my eyes. and of course this is not a best instrument, however, it should be sufficient to measure a deviation of at least 1-2 mm and I can assure you all - the wire was straight enough to make (at least) close to the rest sides. and like you can see and like I said - gcode finished where it supposed to.
I made like was suggested a fake cutter - inserted a pen into collet instead of wire and ran tests at different speeds - everything is fine and it draws a square as it should be...
(I found out by the way that G61.1 gives better result than G61)
which says to me all the machine configurations are correct and mechanical parts are working as they should.
so, I still don't understand why the hell that time it decided to cut only one side in half...

Does anyone think it could be possible that if linuxcnc left paused for prolonged period of time some kind of communication break may occur between comp/linuxcnc and controller card?
it is just my theory attempting to explain what happened...
Any thoughts Gentlemen?

yes, thank you andypugh - I am considering add G4 to my gcode at end of each line.
so, do you think my theory about lost communication between linuxcnc and controller card is valid?
Do you have any suggestions on how that could be fixed?
because I might need to pause the machine and that pause could be long. But I want to be sure this pause will not affect the cutting.

yes, andypugh - parallel port. can it be anything else with linuxcnc?
yes, of course, the power was left on on comp and controller card.
somebody else? alien? (it could be gophers - I have plenty of them on the backyard, but garage door was left locked... )