Ethernet Controller, Part 2

Order PCB arrived. Time for assembly, unfortunately it did not go according to plan:

Tile as a prototype meets the expectations of, but not as ready device!
What went wrong?
The mass of major and minor offenses:

1. The biggest and most disqualifying it vice versa derived one side of the EEPROM - error while drawing a diagram.
2. Places in relays in mirror image, it was possible to work around them soldered on the other side.
3. Raster microcontroller is too exploded, several hundred millimeters on entire length. It can be soldered, but it should be better.
4. In places it is a little too thick and elements overlap.
5. Elements in the housing 1206 are a little too short pads, soldering is more of a nuisance.
6. LED pads are too small.

Okay after constructive criticism see what has succeeded:

Yes, it works! 
Soldering caused some difficulties, but mainly because it was quite monothematic. In total, it helped to refine the solder paste and hot air. Installation took about 2 hours, debugging, unfortunately, all day.
The big problem made to run Ethernet. After checking the continuity of the paths, the connector pinouts, bridged line relays suspects turned out that the guilty are capacitors. By mistake in the diagram, instead of two 56pF were soldered 2k times larger 100nF.
What's next?

I must quickly improve the scheme and the PCB, then already with less uncertainty wait for the next PCB testing in the meantime what is. We must take to heart errors, so as not to repeat them in the future. It is true that nicely worked out?

Other articles about the driver:

• Ethernet driver