Adapting is additive, and each addition costs more than the part. A spacer adds two faces, a sleeve adds two clamps, a transition piece adds two welds, and each of those is a place where a leak can start and a place where the assembly can move. The first adapter is cheap; the fourth is not.
This is the point where owners keep going, because each individual step looks small and the sunk effort is already large. The judgement you need is not whether the next adapter will work, but whether the assembly as a whole is still something you would build if you were starting now.
Count the joints between the two parts you originally wanted to connect. One joint is normal. Two is a design. Three or more is a stack, and a stack behaves differently from a joint: it accumulates tolerance, so the error at the far end is the sum of every interface rather than the error of one.
A stack also accumulates movement. Each interface can move a little, and under vibration those small movements add up at the end of the run, which is why a system with four clamped interfaces rattles where a system with one welded joint does not. Count before you add.
An adapter stops being a spacer when it starts carrying load. If it holds the weight of a section, resists engine movement, or keeps two parts aligned under torque, it has become part of the structure, and it now has to be designed like structure rather than assembled like plumbing.
That is the moment to stop. A structural adapter made from a bought spacer and two clamps will flex, the clamps will fret, and the alignment it was supposed to hold will drift. Structure belongs in welded metal of adequate section, made for the job.
Every adapter takes up length, and length under a car is not free. A stack that adds distance to the run changes where the pipe passes the subframe, the tank and the heat shielding, and the clearance you carefully checked at the start no longer applies at the end.
It also changes what happens at service time. A joint that needs three parts removed before a section can come out is a joint that will not be serviced, and the next repair will happen around it rather than through it. Ask how the assembly comes apart before you decide it goes together.
A clean remake is one part that spans the whole distance with the right fitting at each end. That is exactly the job a drawing-led or sample-led custom run is set up for: the interface is measured once, a mould or a drawing is made to it, and the part arrives with the geometry built in rather than assembled on the car.
The custom work process exists for this case, and it is worth using before the adaptation stack reaches three interfaces, because a part made to the measured interface removes every joint the stack was adding. The result is fewer seals, fewer clamps and one thing to check instead of five.
Remake when any of these is true: the stack has reached three interfaces, an adapter is carrying load, the stack has changed clearance anywhere along the run, or the assembly cannot be taken apart without removing something else. Any one of those makes the remake the cheaper route over the life of the car.
Keep adapting when the change is one interface, nothing structural, clearance unchanged, and the adapter is a standard welded fabrication part that can be replaced on its own. That is a repair. Beyond it, you are building a prototype on the car, and prototypes belong on a bench.
