In my opinion and experience, LithiumFE (Li-FeS2) primary cells are the longest lasting batteries, over the widest temperature range, for this kind of application. Their initial charge is about 1.75V, it stablizes to about 1.7, and maintains this voltage until very close to end of life, dropping off the cliff at about 1.5 volts. Hence, two of these, either AAA or AA sized will give you useful voltage for their entire life and their shelf life is 20 years (if you believe Eveready - I personally haven't experienced this - yet). No need for an LDO and it's HIGH quiescent current
Also, once you have sufficient voltage to operate, throwing more batteries at the problem actually reduces battery life. It's not about volts, it's about mAH and the mA load will increase with increased voltage while the mAH of each battery remains the same. Net, 4 batteries forcing you to use an LDO will die sooner than the same 2 batteries where you don't need an LDO.
Regarding sensors, processors, and radios, most these days will operate well into 2V, quite a few down to 1.8V reliably. The processor and RFM69 will operate down to 1.8V, I know this from personal experience and data from JoeLucid. However, the ESR of the power source must be able to maintain voltage no less than 1.8V throughout operation.
If you use an external timer/WDT chip such as TPL5110 or TPL5010 to control sample sequencing, you can easily get your 'sleeping' current into sub-100nA region.
Given this, your goal of 5+ years is certainly achievable, with careful design practices... Your greatest risk of failure is not from design currents draining your batteries, but corrosion causing leakage and unpredicted failures.
Tom
Update: added comment about extra batteries.