A600 glitches after recap

[Castellen]

Sometimes multi-layer ceramic capacitors can go resistive if they've been subject to physical stress (cracking) or contamination such as corrosion damage from capacitor electrolyte.  It's not uncommon for this kind of fault.  If the capacitor appears damaged/corroded or it there's any sign of heating, it should be replaced.

The valid clue here is that the 'insert disk' animation is getting screwed up.  The fact that the rest of the graphics mainly appear to be drawn OK suggest that the graphics output hardware (U4, U12, etc) is likely OK.

I'm not sure exactly how the disk animation works at the software level, but in hardware it relies on the level 3 interrupt (_INT3 on the schematic).  This is generated by U2/Agnus, which gets passed to the interrupt handler U3/Paula; presumably the software is relying on the interrupt to move the disk image data between different areas of memory to create the animation, not sure.  Based on the clues in the screen photo, something about this interrupt is broken.

While viewing the 'insert disk' screen, monitor what's happening with _INT3.  The line should be inactive all the time (high), and only strobes active (low) while the disk is 'moving', with an active duration of around 200nSec.  Make sure this signal is arriving at U3/Paula pin 20.  If that appears to be OK, there's possibly something broken with the interrupt priority level lines from U3/Paula and the 68000, so check continuity of the three IPL lines between U1 and U3.

[Castellen]

Also, if I power on the machine with the probe on _INT3 most of time the screen remains white. More rarely it shows "Software failure" like in the attached picture.

That sounds suspicious.  Since _INT3 is actively driven by a low impedance output from U2, connecting a scope probe (typically 10M Ohm impedance) shouldn't make any difference.  Sounds a lot like _INT3 might be open-circuit/floating, but if you've confirmed continuity, then the problem may lay elsewhere.  It's also possible that some of the graphics operations use DMA modes, so perhaps pay attention to the DMAL signal.  That should have activity all of the time (groups of 4 x active high strobes), as generated by U3.

and removed Paula to check traces under the chip (some of them were "blacky") I putted all back and powered on the A600 with some ugly glitches on the half-down screen like in attached pictures.

I gather the A600 was originally working correctly (no graphics display issues) before you did anything.  And it was OK after replacing the capacitors?  But the problems appeared immediately after reworking U3/Paula?  If so, it's very likely that the fault is a direct result of an issue you've introduced; such as a solder short between two lines, intermittent/unsoldered joint, damaged PCB track/pad, etc.  As opposed to reverse engineering the system by forum posts to track down a potentially very complex issue, my suggestion would be to very carefully go over everything you've done (again), inspect the entire board for a stray blob of solder, etc.

I don't know what your rework process was (i.e. temperature control, etc), hence there could be a possibility that U3 has been damaged.  Though I'd be looking for a connection issue or short between lines to begin with.

[Castellen]

I would suspect, in case, of bad or little highter impedence on vias?

Vias should be a normal part of the circuit, so will have a continuity close to 0.1 Ohms.  Anything much higher means the via is open circuit.

Don't rule out the possibility of an accidental solder short between two lines.  This can be harder to find, because the end-to-end continuity will measure OK, but it will also be joined to some other unknown line.  You can usually see this condition on the scope, since if two logic outputs are incorrectly joined, you'll see 'halfway' logic levels as opposed to the usual 0v to 4-5V waveform.

Solder joints on the PLCC look OK.  Rework temperature of 380°C is good.  Could have also been a coincidence that a weakened via somewhere has been damaged while you were performing the rework on U3.  Anything is a possibility when there's been corrosion damage.

Graphics issues could also be caused by a fault with U2/Agnus, as that performs the graphics data move operations during the 'insert disk' animation.  Though I don't see how that would have failed from what you were doing.

The fastest way to progress at this point would be to get a working A500, A2000 or another A600; power the good working machine up on the 'insert disk' screen, and do the same with this problem A600.  Doesn't have to be connected to a monitor.  Use the scope to compare each pin on U2/Agnus and U3/Paula between the good and problem machines to identify any signal differences while viewing the 'insert disk' animation on both machines, especially noting specific signal activity while the disk animation is moving.  Note there are some pin number differences, A500/A2000 have the 8364 in DIP48 package, the 8364 in A600 is PLCC44.  There are also a few minor pin function differences between the 8371/8372 in A500/A2000 and the 8375 in A600.