I keep seeing the same two oscillator selection mistakes in design reviews

After a few too many design reviews, I've noticed the exact same two mistakes come up over and over again:

1. The "just in case" over-spec Someone will put a $150+ OCXO into a design that only needs ±2ppm stability. Then they wonder why their battery life dropped from 5 years to 6 months, and why the board needs a 5-minute warm-up before it's stable. The thing is: TCXOs are good enough for most consumer and industrial use cases. ±0.5 to ±2.5ppm over temperature covers 90% of designs. You don't need an oven unless you actually need sub-50ppb stability.
2. The "we'll just fix it in software" under-spec On the flip side: someone uses a basic XO or cheap TCXO as the reference for a 1GSPS ADC, then spends weeks debugging why the SNR is 10dB off. Clock phase noise at 10kHz offset doesn't care about your software filters. Or the classic: picking a TCXO for a 5G small cell that actually needs Stratum 3E stability (±50ppb), because someone looked at the ±2ppm spec and thought "that's fine, right?" My lazy decision tree after 10 years of this: Need to pull the frequency with a control voltage for a PLL? → VCXO. Need better than ±50ppb over -40 to +85°C? → OCXO. Need to hold time without GNSS for more than a few hours? → OCXO. Literally everything else? → TCXO. Stop overthinking it. The other thing nobody talks about: warm-up time. A lot of engineers forget that OCXOs can take 20 seconds to 10 minutes to reach stable operating temperature. If your device has to boot and transmit in <1 second, an OCXO is going to give you problems. What's the most common oscillator mistake you see in designs? Is it the over-spec "just in case" crowd, or the under-spec "software will fix it" crowd?