Only you and the environmental authorities can answer this question. A truly benign fluid can be dumped onto the floor! You can pipe it to a close process sewer if it is not too hot. You can direct it into a small drum for later disposal if this will be a rare relief. There are countless possibilities but again, only you can determine this.

Some companies will cascade them across the block valves.
Suppose that one has a block valve at the offsite battery limit (B/L) and one at an onsite B/L block valve. They will place a thermal relief across the onsite block valve and another at the offsite block valve. The problem becomes a bit hairy because one has to keep setting the downstream valves at a lower setting. In addition if one were to take the downstream tank out of service then the upstream block valves have no place to discharge to.
It is done but it creates a lot of additional controls.

A balanced or bellows style TRV are ideal for cascading relief system.  The TRV set point is not referenced to downstream pressure.  As such, it is not difficult to design provided you spec the correct PRV.  Look at Swagelok R31, Anderson Greenwood 81 series, etc. Note, the R31 are not ASME BPV listed, the 81 series are.  Cascading problems arise when trying to use a PRV that references downstream pressure for the relief setting.

CRG
Correct and good point.
But bellows type valves bring on their own set of problems even when delta P is not a problem.

Bellows wear out.
The bonnet has an opening to the atmosphere.

One still has to manage the downstream blinds in an offsite area.

In an offsite area a leaking bellows is less likely to be noticed than in an onsite area. But they are a means of removing the variable downstream pressure from the set pressure.

CRG

Do you mean that using a balanced bellow type of thermal relief valve in a cascading relief system, you only have to specify set pressure and there is no need to specify  back pressure (variable or constant)?  Do you need to specify spring setting (differential set pressure)? A cacading relief system is exactly what we need for the OSBL piping system and I have a problem to properly prepare a releif valve datasheet. Could you list the required data for a complete thermal relief valve? Is it better to use a pilot-operated thermal relief valve in a cascading relief system? I will appreciate it very much for your guidance.

Note: There is a difference between bellows type and balanced relief valves.  Anderson Greenwood 81P and Swagelok R31 do not use a bellows (see http://www.andersongreenwood.com/direct_spring.asp for data on 81P).  However, there is a seal between the stem and atmosphere with a balanced valve; as such, a potential for a leak is possible.  To put this in perspective, the block valve which is creating the pressure barrier necessitating the TRV (thermal relief valve) has a seal separating the fluid from the atmosphere as well.  For information on one type of ASME BPV listed bellows TRV, look at Hydroseals section H,  http://www.hydroseal.com/Products/h.html .  

Check with supplier/manufacturer to verify that their TRV models are fully balanced if your design requires this.

As I recall, B31.4 does not require ASME BPV listed PRVs and B31.3 does.  Not positive about this without looking it up so make sure you follow the design/code requirements.  Consequently, many owners prefer to use only the listed PRVs for B31.4 designs.  I agree with them in most cases.

Back pressure is relevant for sizing the orifice size in a fully balanced TRV.  

Provide supplier with set pressure, downstream pressure, fluid type, temperature range of fluid, relief flow requirements, and type of connection.  Most suppliers of TRVs have information for sizing their valves on the Web.  Use this information to ensure that their recommendations for TRVs meet your requirements.  

Make sure all scenarios are evaluated and discussed with owner, because the installations of one blind flange installed downstream for service work may disable the required thermal protection for the entire pipeline.