Hardening a GNSS front end is a layering problem, because no single LNA choice fixes both sensitivity and jamming at once. Start by defining the worst jammer you must survive, not the typical one: an in-band continuous-wave jammer a few MHz away behaves very differently from a pulsed or broadband jammer, and the filter and linearity budget follow from that.
Put a narrowband SAW or BAW filter ahead of the LNA, as close to the antenna as the layout allows. That filter is what keeps out-of-band cellular, Wi-Fi and walkie-talkie harmonics from driving the LNA into compression. Accept the 1-2 dB insertion loss it adds, and recover sensitivity downstream by choosing an LNA with a noise figure in the 0.8-1.5 dB range and an out-of-band IIP3 high enough that the filtered jammer still does not compress it. Chasing a 0.5 dB noise figure at the cost of 10 dB of IIP3 is the usual mistake here.
Add a limiter or PIN-diode protection stage only if the field exposure can reach damaging levels, and place it after the pre-filter so it does not add loss in normal operation. Follow the LNA with a second, tighter filter before the mixer or ADC, and use AGC with a fast attack so a strong jammer does not saturate the baseband.
On the bench, measure C/N0 as a function of jammer power for each jammer type, with the antenna in place, and repeat with the antenna replaced by a terminated load to separate front-end compression from antenna effects. Also check power integrity, as George notes: a jammer can modulate a weak LDO or a shared supply and look like front-end desensitisation. A clean supply, a solid ground return under the LNA, and keeping the GNSS antenna away from vehicle transmitters often recover more C/N0 than another 0.2 dB of noise figure.