Lawrence Berkeley National Lab here. I hope this response is not too long.

We have adapted the Fermilab methodology for ODH risk analysis to suit our purposes. You can read our policy on cryogenic liquids here, though be aware that we are looking to update this soon with information from some additional lessons learned. But that's only in relation to cryo burns, not oxygen deficiency.

Following the Fermilab model, we calculate the oxygen deficiency hazard class of the room. If it's a very simple case of cryo storage, filling small Dewars, and general use for vacuum traps and freezing samples and such, then we generally start with a simplified calculator that only looks at total storage, largest transfer Dewar used, and largest amount used openly. It doesn't account for ventilation or anything. If that comes back with ODH class 0, then we know they're fine because the model is overly conservative by not accounting for ventilation and assuming 1 ACH (all of our laboratory spaces are required to have at least 6ACH and most are in the 20s). If it's ODH class 1 or greater, we move onto a more detailed analysis.

If there's more going on than described above (examples include constant flow of cryo through a flow cryostat, custom built or modified systems, large cryo baths, work being performed in a cleanroom or coldroom or other space with reduced fresh air ventilation, etc.) then we go straight to a more detailed analysis. We measure the fresh air ventilation rate of the room, ensure our dimensions are accurate, and attempt to account for all possible failure scenarios. Fermilab has lists of failure rates for common parts and systems that we draw from. We also look at the specifications of valves and such to get the flow factors and calculate the maximum release rates based on the pressure of the system. We attempt to get release rates for all routine work and failure scenarios as best we can, and calculate the lowest oxygen concentration. We often apply a safety factor of 2 or more for routine work.

If the ODH class of the room is 1 or greater, we first look for ways that we can reduce that to 0. That may involve asking the researchers to move their work into another room that is larger or has better ventilation, or it may involve reducing quantities used. If we can't get to the ODH class below 1, then we require a permanent oxygen monitor to be installed, along with other controls outlined in our policy. At the moment, we do not have any work being done at ODH 2 or greater, as this would require some serious scrutiny before we were comfortable with it. The vast majority of our work is ODH 0.

We do not allow any work where routine scenarios could put the room into oxygen deficiency (< 19.5%). If the analysis indicates that routine work could drop the oxygen anywhere close to 19.5% (in practice anything 20.5% or less), we also follow up by monitoring the work the first time it is done with multiple data logging four-gas meters in the room to verify that we are not in danger of creating an oxygen deficient work environment. So far, every time I've done a verification, it has turned out that the actual oxygen concentration in the room is higher than what was calculated. We also like to do this verification for any indoor Dewar filling operations, regardless of what the calculations show.

Permanent oxygen deficiency monitors are expensive investments and require a lot of maintenance to keep them calibrated and ensure they're working properly. Our researchers generally prefer to try to move or modify the work to achieve ODH 0 before resorting to them.