I've been working to get my MMU2S working on my MK2.5/S. I use mostly PETG and mostly Amazonbasics or eSun. Single color prints work just fine. Lots of issues with loading failures.
First issue I noticed was that very often the filament would block trying to be loaded into the extruder. I noticed the tubing between the MMU and extruder was curved/bent near the extruder sometimes depending on the extruder location. If I held the PTFE tubing during such a load attempt to make it straighter, the filament would be loaded ok. So I designed this hack ( https://www.prusaprinters.org/prints/4227-mk25s-extruder-tubing-support) to slide over the festo connector on the extruder and straighten out the tubing near the extruder some more. This helped quite a bit to eliminate those errors. Now I haven't seen one in many testing attempts.
Second, I suspected there was too much friction with the normal PTFE tubing and I couldn't easily see the filament to see what shape the tips were, so I replaced the tubing with 2.5mm ID clear tubing that seems to also help. I have not replaced the small PTFE tubes in the MMU itself. That hasn't been a problem so far, but I may do it at some point. I further replaced the tubing between the mmu and extruder with 3mm ID tubing, but can't be sure if that change was critical or not.
Third, I was now getting some load failures with bad IR sensor detection on the filament load. So I replaced the standard extruder idler with this one:
And fed a piece of filament manually, and adjusted the bolt to barely indicate a "1" on the IR sensor reading when the filament was present. And then screwed in the bolt maybe another 1/2 turn. I wanted to avoid having the sensor detect long/skinny tip tails. I have noticed I get a lot of variation in tips. Often I get really nice shaped tips with no tails. Then I will see long skinny (hairlike) tails, sometimes up to maybe 4cm! On the same filament, on the same print.
With those changes, I was able to run a test print for about 3.5 hours: This one https://www.thingiverse.com/thing:3120233 (small 5 blocks). Then I got a load failure. This time I noticed the filament seemed to have been pushed too far back in the MMU and the MMU wasn't able to pull the filament back in on a subsequent load attempt. I was able to recover from the load failure after I simply pushed the filament a little back into the MMU, all was fine. I suspect a long tail caused the MMU to retract the filament too far. I think the FINDA needs to be adjusted "outward" some more. Similar to the adjustment on the IR sensor. (Next adjustment to try).
Finally, I noticed the wipe tower are really "junky". Especially on the first few layers, when the wipe tower was built with bridge supports (due to not having a tool change for some filaments on that layer), the plastic would not quite extrude cleanly between the supports. And it would leave plastic blobbed near the support cross lines. This would lead to more junk later. And a few times to a layer shift when the hotend would get stuck on the built up junk. This was happening when I was printing at 225 degrees. When I increased the temp to 235/238, everything was much better. I still have a little extra extrusion where the extruder stops moving during a tool change. Will have to dig into that some more later.
I also have adjusted the PrusaSlicer ramming profile to be just 1s and the extruder unload to be 130mm, but I am not sure that is important with all the above changes. I first tried those changes when I noticed that my tips on simple load/unload of a filament during testing or a single color print looked just fine. But during multicolor prints, they had really long tails and tips. Since ramming isn't done (I think) for single prints or just an unload, I reasoned that ramming might be causing issues. These changes seemed to improve the tip shape, but now I am not so sure. With all the above changes, I am also not so sure now. I need to go back and use stock settings and see if prints still work.
Anyway, maybe the above will help someone else get their MMU2/S working better.