The factor to focus on is what your plans are for the site itself into the future. If you plan (or think you will need to) develop the content or operation of the site in the future (e.g. adding new things / services not currently provided) then it is worth considering upgrading to EE3, otherwise the best course is to stay with EE2.
Why? Well it is clear that EllisLabs and the EE hard-core development community are moving on to EE3 - although EE2's EOL has been delayed (due to just the problems with upgrading you outline) EllisLabs will stop maintaining EE2 soon, and there is little or no development going on for EE2 add-ons, which are either being abandoned or migrated to EE3 versions.
The EE2 system you have now, while it will continue to work 'as-is', will get harder to develop: you'll be stuck with your current version of php, and with the set of Add-ons you have now. You will be able to re-skin the front end of the site and keep that up to date, but the underlying data model will be limited to being something similar to what you have today.
Upgrading to EE3 is non-trivial (outside of the "no add-ons used" edge case), but would appear to be essential for a site that will need back-end development in years to come.
There are now EE3 alternatives to most of the widely used EE2 add-ons that haven't got EE3 versions - though replacing them is not always easy. Others are still in the pipeline (e.g. Assets has just released an EE3 beta..., Publisher is '2-4 weeks' away from releasing a beta). So while it was almost impossible to migrate some sites a few months ago, it would seem that by early 2017 most of the obstacles will have been removed. There will be some add-ons that will regrettably remain absent- CE_Cache being one of them, though (if you can get your brain around it) it looks like you can get a similar result using the caching features of
Tough call though. Hope these thoughts useful etc.