Switching to fixed window size increased the risk of failure

During initial implementation stage, all auto tests were running from my laptop. I used max window and never had an issue with the appearance of various elements on the screen.

Now, we got to the stage where auto test suite will be executed by more than one person, so that we have to consider difference in screen resolutions, OS (MAC/Windows), etc.

To get on the same page, we decided to use a fixed window size instead of max window:

'chromeOptions': { args: ['--no-sandbox', '--window-size=1366,768'] } },

Once fixed window size is in place, I started observing slight inconsistencies with the display. The same window could appear slightly above or below the "expected" location. For most parts, it doesn't affect outcome of test runs... but about 10% of the time it leads to failure due to "element is not visible" or "element is not clickable" error.

In summary, switching to fixed window size appeared to increase probability of a failure by 10%, which will result a lot of extraneous "noise" when everyone will start using this suite.

Is there a way to achieve the same consistency as max window method, or this is a known imperfection/limitation on protractor side?