Is there any tool that can do this? Re-convert the .dat files into static files?

We want to rollback our map to previous point, but we are facing the problem of having lost the statics... Anyone has any ideas?

You have two choices:
Write an converter based on the uoconvert sources with a language of your choice. Or you use escript functionalities: via getmapinfo and liststaticsatlocation you get all the needed information's and via openbinaryfile function you can write the corresponding mul files.
At least I do not know any existing tool which does the job for you.

Turley wrote:Write an converter based on the uoconvert sources with a language of your choice.

This was what I was thinking. I even took a peak into uoconvert source code, but I think my knowledge of UO and C++ are kinda basic.

Turley wrote:Or you use escript functionalities: via getmapinfo and liststaticsatlocation you get all the needed information's and via openbinaryfile function you can write the corresponding mul files.

Could you be more specific? If you workout an example I'm sure I'll be able to write the thing. My main problem is that I don't know how I should compose the binary file to match the several static files. Although, in this matter I would rather write a program that could benefit others.

a classic lecture is also
http://uo.stratics.com/heptazane/fileformats.shtml#3.8
you can also check out the fiddler sources which are also available at github.

the easiest mul file is the map.mul. it just contains graphic and z value.
the file consists of (width/8 * height/8) blocks. each block has a unknown int as header and then the 64entries
thus to write the mul its just:

for (xblock<width/8; ++xblock)
for (yblock<height/8;++yblock)
mul.write((int)0)
for (xi<8) for (yi<8) // write the 64 entries
mul.write(graphic of xi/yi)
mul.write(z of xi/yi)

I won't promise I'll write this tool, but I'll look into it.

Please refresh my memory, what UOconvert does? Manage static files?

It transforms the statics into something POL can read, if I'm not mistaken...