An indirect answer to your question.
We have done 0.5mm pitch LQFP DIY boards.
We used 'proper' laser printable PCB Artwork Drafting film
We spent quite a lot of time calibrating the exposure time of the PCB in the UV box, and IIRC it was sensitive to a +/- 5 second variation. Too short or too long produced poor results. IIRC, we made a 'PCB' with some test patterns for different track/space distances to help us calibrate things.
We still had quite a lot of trouble getting good results. Then we discovered that the laser printer was stepping in, and trying to produce a 'grayscale' when the PCB artwork 'pixel' boundaries didn't match its own idea of pixel boundaries. When we examined the artwork under a microscope, we could see that edges were defined by a fuzzy (dithered) half-tone pattern, rather than a much denser, more uniform edge.
We improved the results by 'fiddling around' with printer settings.
Then I redid the footprint of the 0.5mm pitch LQFP part so that the gap between pads was slightly bigger. That gave better results.
Edit:
I know folks who have tried a 1,700GBP 'ebay' PCB mill. AFAIK they gave up due to difficulties getting consistent results. They have now spent a lot more to get a proper LPKF milling machine.
Edit2:
Is the entire board 'packed', with a need for 0.25mm track/space everywhere or is it mainly around the 80pin part?
Depending on where you are in your development process, and the sort of issues you are needing to fix, a way to reduce the pain might be to make a 'breakout' board for the LQFP part with your high-quality manufacturer. That would have lead-time, but once you have some, you might be able to turn-round the rest of the PCB using DIY.
It may also be the breakout PCB can solve some of your layout issues. If you put it's decoupling capacitors etc on its breakout PCB, its behaviour might be okay. My experience is manufactured vias are much smaller than DIY vias freeing up board area. Further, putting vias under the chip are awkward to do on a DIY PCB. So you might get a lot of benefit from the manufactured breakout, and hence make the remaining DIY PCB easier to route.
A traditional breakout usually has pins on 0.1" centres, in a square around the chip. You don't need to do that. You could use finer pitch connections and with pins in a non-rectangular, convenient, shape for your problem.
Maybe even consider doing a 4-layer breakout PCB, to make the rest of the board as simple to layout and make as practical.
Best Answer
Some key things to look for:
Delivery time. Many low cost fabs require several weeks to schedule and build your boards. Three-day turn-arounds cost much more.
Will they respect your requirements (fab notes), or just build to a standard set of tolerances and specs? Many low cost vendors severely limit what they will accept in fab notes.
Quality. Do you trust this shop to build your board right the first time, or is there a chance they'll mess it up and have to re-do it, causing a delay?
Lower-level materials. Are they buying the actual laminate materials from high quality vendors or just whatever's cheapest at the moment. (Does your application need the higher-quality material?) Will they use the same material for every lot? If they're just buying at lowest cost, the product is likely to vary from lot to lot.
Support. Are they providing engineering support to review your design and catch mistakes (mostly your layout mistakes) before you spend money on the fab.
Test. Are they providing 100% connectivity testing on the boards after manufacturing?
Certifications. Can the provide UL 94V-0 fire resistance certification on your boards?
Capacity. When your prototype works and you're ready to build 1000, 10,000, or 100,000 boards per month, can they support you?
Capability. Do you need 3/3 space/trace, gold plating, thin dielectrics, impedance control, microvias, etc., etc? Higher-technology designs need higher-cost equipment to build and more attention to detail when building them.
Obviously it depends on your project which of these qualities are worth paying extra for.
Traditional tin/lead is easier to work with for hand assembly, but can't be legally used in products you want to sell in Europe, China and probably some other countries.
Pure tin coating is a reasonable alternative to tin/lead if you can't use tin/lead.
Another option is organic solderability preservative (OSP) over bare copper. This is common for volume production.
Gold is useful for corrosion resistance if you are making an edgecard connector, or if you will be wirebonding to a chip-on-board component. Different gold-plating processes are generally used in these two situations. If you need to know the difference, it's worth opening a separate question about it.
If you are hand-assembling your boards and in the USA, tin-lead is probably your best choice.
Measuring the quality of PCBs is pretty involved. Most defects happen when you try to push to the limits of technology (3 mil traces, 6 mil vias, 0.5-mm BGA pads, etc). If you are only working at low volumes and with forgiving design rules (8/8 or larger), you aren't likely to run into quality issues, even with low-cost vendors.