raj_pm is right about loaded vs static — that's the measurement most shops miss. We actually found a 0.04mm difference between static and loaded parallelism on our 300T, which explained why parts were fine on light jobs but went out of spec on heavy coining work.
On the measurement method: we switched from dial indicators to a Renishaw XL-80 laser for parallelism checks about 2 years ago. The laser is faster and more repeatable, but honestly for most shops the dial indicator method works fine if you're careful about your setup. The biggest source of error with dial indicators isn't the indicator itself — it's the magnetic base shifting on the bolster surface. We started using a dedicated fixture plate bolted to the bolster and our measurement repeatability went from ±0.01mm to ±0.003mm overnight.
Frequency: we check monthly on our high-precision press (medical parts) and quarterly on the others. If you're seeing a trend — like parallelism degrading 0.005mm per month — that's gib wear and you should plan for adjustment before it affects parts.