Crystal Oscilator Circuit Problems

[floobydust]

It's misleading that it worked on a breadboard with different (through-hole) parts - the IC, crystal, caps. Everything has been changed, so back to square #1.
Maybe post pics of your PCB layout we can look for issues there.
A breadboard's parasitic capacitance, I measure 3pF between two strips and higher if mounted on a metal plate or workbench. Also the high hum and noise, common-mode depending on if the power supply was earth-grounded, can help a weak oscillator start up. Physically smaller crystals have higher ESR as well.

Strange HEF4060B datasheet, it doesn't have Q14 on pin 3, or the tri-state on the gate. Sometimes SMT packaged IC's are different than their through-hole version.

Lots of posts on the web using CD4060 for nixie clocks and such https://www.eevblog.com/forum/beginners/32-768khz-pierce-crystal-oscillator-for-nixie-clock/ as well.

[james_s]

The capacitors could easily be off by an order of magnitude or more and still be well below what you can measure with something like a multimeter. It's not unheard of for something like that to happen, last year I had to debug some prototypes built by a company and I found several capacitors that were way off, like 1uF where it should have had 15pf caps on the clock crystal.

[sarahMCML]

For some strange reason, the HEF4060B datasheet mark's each flipflop's output 1 less than than the value of the latch it's actually connected to. Weird!

[james_s]

The flipflops are numbered starting from 1. The outputs are numbered starting from 0, as is convention.

[floobydust]

CD4060B TI/RCA/Harris and NS datasheet Q4-Q10, Q12-Q14.
NXP HEF4060B datasheet Q3-Q9, Q11-Q13. Just pointing out the different numbering convention, which is annoying.
NXP also includes this graph for the osc. amp transconductance, VDD affects it almost 3:1.

[bson]

One thing to keep in mind with small, extremely low-power SMD crystals is that flux bridges between the pads can be source of trouble.

Don't ask how I know. 

[Benta]

I would say good luck getting it to start with the series resistance value you recommend 1MEG 

How about grabbing a simple calculator and compute what 0.1 uW results in as series resistance at the voltages involved? THEN you can come back with your smirking "smiley".
You've obviously never worked with these kind of crystals before.

[floobydust]

Sorry we're not all masters like yourself. 1MEG with say 10pF load seems excessive  over industrial temperatures it would have to be a precise design.
Majority of these oscillators are datasheet copies, to dig in is very difficult. I have designed these oscillators too but the problem as I have mentioned the Observer effect makes measurements nearly impossible, unless a low cap JFET probe or inductive pickup which watchmakers use to calibrate frequency. LT Spice simulations are not useful either because the parasitic capacitances are not really known nor the crystal's SPICE model. There is always empirical work I find, and I run out of project time and patience.
One product's 32kHz oscillator refused to startup around 5°C but colder or warmer by a few degrees it would reliably start, it was awful.