The gap between disconnecting the old chiller and starting up the new one is where downtime risk concentrates - plan that window deliberately.
Confirm the new unit's electrical and piping connections match (or have been adapted to match) the existing infrastructure before the old chiller is disconnected - re-terminating a mismatched electrical service or re-routing piping after the old unit is already out extends downtime that could have been avoided with a pre-check. If the new unit's footprint or connection points differ from the old one, that adaptation work should happen before, not during, the swap.
The old unit's refrigerant needs to be recovered by a licensed technician before disposal, regardless of refrigerant type - this is a regulatory requirement, not optional, and it's worth scheduling as its own step rather than assuming it happens automatically during removal. For an R-22 unit specifically, recovered refrigerant may have real reclaim value, which is worth asking your service contractor about rather than treating it as waste.
Plan the actual downtime window realistically: disconnection, rigging the old unit out, positioning and connecting the new one, and startup/commissioning all take real time, and a process that can't tolerate an extended cooling gap may need a temporary bridge chiller (arranged by the site or its contractor - see Emergency & Backup Cooling for Critical Spaces under Medical & Healthcare for the connection planning) for the changeover window.
Confirm electrical, piping, pad footprint and rigging access while the old chiller is still running. A mismatch found mid-swap costs the most, because the process is already offline and waiting.
Most swap-out delays trace back to one of two causes: a connection mismatch discovered after the old unit is already disconnected, or a rigging access problem that wasn't scoped before the crew showed up. Both are avoidable with the pre-check step, and both are expensive precisely because they're discovered mid-swap, when the process is already offline and waiting.
The second most common delay is controls integration - if the new unit ties into a building management system or a process controller the old one didn't, budget extra commissioning time for that specifically, rather than assuming startup covers it automatically.