Sigh. I fear what we have now is a “too many cooks in the kitchen” problem here, and now you are getting tossed to and fro as if by wind.
All of these suggestions about things for you to try were being thrown around here by others because they didn’t trust that you had actually tried two different RB3011s side-by-side, and consistently would get two different results (one always works, one always doesn’t).
So before we go off getting distracted by a bunch of unnecessary other things, let us first establish THAT fact, to the satisfaction of everybody here.
Let’s all take a step back and just employ some freaking common sense. Obviously if one RB3011 works, but you haven’t actually tried to Netinstall to that one again for, like, 6 months, and you are just trusting that it still works because it worked last time you tried it, well, that doesn’t cut it. Something could have happened either to your PC or to your network in the meantime. Also obviously, if one RB3011 (still) works, the other one doesn’t, but there are any differences either between their software and/or firmware versions, or the PC host you are running Netinstall from, or the network you are attaching them to, then logic dictates that any of those differences could potentially be the factor, and not the RB3011 itself. So we need to eliminate any of these differences, so that there is no question we are talking about a fair and direct apples-to-apples comparison here.
It seems to me that there is no reason why, rather than focusing all of your attention on the “bad” one, you can’t have both “good” and “bad” sitting on your desk to play with at the same time, and switch between them during your tests. This would rather unequivocally demonstrate that the problem is somehow with one of the routers, and not with anything else that you are doing or failing to account for. You try one of them one minute, and it fails to work. You try another one the very next minute, and it works fine. You haven’t done anything different in the very short span of time between testing both of them (not even PC reboots). The results of such a test would be fairly clear-cut and unambiguous.
So. At this point, you have demonstrated that the “bad” RB3011 is still “bad”. Now, do the exact same things to the “good” one (downgrading it back to 7.6 or whatever, both RouterOS and RouterBOOT), and attempt a Netinstall to THAT one, using the exact same PC, with the exact same IP addresses, with the exact same version of Netinstall running on that PC, and with the “good” 3011 plugged into the network in the exact same way that the “bad” one was when it failed yet again.
Within minutes of each other, run two tests where you change ZERO variables, EXCEPT for ONE: the router itself. This is Troubleshooting 101. I read your past posts and came away believing you had already done this, but others here clearly still doubt, and to be fair it is a little ambiguous. So let’s leave ZERO doubt about this, so we can move on and stop wasting time on things of no consequence.
It didn’t need to be 7.6. They just both need to be unquestionably THE SAME, regardless of what that “the same” is. Because if one works with a given version of software, and the other one doesn’t with THE EXACT SAME VERSION, and the results are always 100% reproducible, then it’s really really hard to argue that the software is in any way, shape, or form the cause of the problem. So, no, you did not need to downgrade to 7.6 if they were both already running 7.18.2. You just needed to test both back-to-back on the same PC + network at that point, to establish there is a clear difference between two physically distinct devices of the identical model, with both running identical software.
To this point, I asked last time for you to run the following command on BOTH the working and non-working routers, and to paste the whole output here from each (with each one clearly labeled as to whether it belongs to the “good” or the “bad” one):
/system routerboard print
This will highlight not only whether they both are actually running the exact same software and firmware, but also whether there is any possible hardware difference between the two (one possibly a later revision than the other, for example?).
That’s what he means, yes, just like you would with an .npk. Skip this, at least or now. All that this is for is for upgrading the backup/factory bootloader. You said you had already done this. If true, then the output of the command I gave above will prove this unquestionably to be the case.