The Expensive RAM Optimization

An async image-import worker consumed significant memory.

The local technical answer was plausible: rewrite it in a runtime with better concurrency and lower memory overhead.

The broader cost was larger than the infrastructure problem:

• several engineers spent months rewriting;
• the team had no prior production experience with the new stack;
• the old implementation still needed support during migration;
• infrastructure savings were small compared with engineering cost and opportunity cost.

total technical cost =
infrastructure
+ engineering time
+ delivery delay
+ operational burden
+ opportunity cost

The lesson is not “never rewrite”. The lesson is that a technically better runtime still has to earn its organizational cost.